path
stringlengths 7
265
| concatenated_notebook
stringlengths 46
17M
|
|---|---|
source/2021/500Answer/content/019_can_build_horizontal_rectangle.ipynb | ###Markdown
第19讲 重新认识矩形 Assignments 作业 1. 熟悉`prepare_paper`方法：导入`qianglib.py`，调用`prepare_paper`方法，修改该方法接受的参数值，使得`scale`值分别为10和50，min_x分别为0，10，min_y分别为0, 20，组合不同的这些参数值调用该方法。观察绘制出的方格值有什么变化，从中试图小结个该方法这些参数的意义。 Your Answer: 1. scale的意义: 2. min_x的意义: 3. min_y的意义:
###Code
# 你需要预先从turtle和qianglib库中导入需要的方法
weight, height = 800, 600
setup(weight, height, 0, 0)
prepare_paper(width, height, scale=10, min_x=0, min_y=0)
prepare_paper(width, height, scale=10, min_x=0, min_y=20, max_y=25)
prepare_paper(width, height, scale=10, min_x=10, min_y=0, max_y=25)
prepare_paper(width, height, scale=10, min_x=10, min_y=20, max_y=25)
prepare_paper(width, height, scale=50, min_x=0, min_y=0, max_y=25)
prepare_paper(width, height, scale=50, min_x=0, min_y=20, max_y=25)
prepare_paper(width, height, scale=50, min_x=10, min_y=0, max_y=25)
prepare_paper(width, height, scale=50, min_x=10, min_y=20, max_y=25)
###Output
_____no_output_____
###Markdown
2. 使用本讲示例的代码调出绘图纸。下面以供给出了5组数据，每一组数据包括两个点的坐标。 对于每一组数据： 1. 请在绘图纸上使用`qianglib`中的`mark`方法标记这两个点， 2. 用`qianglib`方法中的`line`方法连接这两个点成一条线段。 3. 手工找到线段的中点，用`mark`方法标记该中点。 以端点坐标表示的线段 1. (0, 0)和(20, 0) 2. (0, 0)和(0,16) 3. (0, 0)和(20, 16) 4. (10, 20)和(30, 20) 5. (10, 20)和(35, 25)
###Code
# A.
points = [(0, 0), (20, 0)]
for point in points:
mark(point, info=str(point))
center = (10, 0)
mark(center, info=str(center))
# B.
# E.
points = [(10, 20), (35, 25)]
for point in points:
mark(point, info=str(point))
center = (25, 22)
mark(center, info=str(center))
line(points[0], center)
line(center, points[1])
line(points[0], points[1])
###Output
_____no_output_____
###Markdown
3. 以上一题为参考，编写一个方法专门计算一条线段中点的坐标，这条线段以两个点的坐标的形式给出，方法返回线段中点的坐标。该方法的定义如下，请完整的实现该方法。
###Code
def line_center(point1, point2): # point1, poit2 format of (x1, y1)
# center = None
# TODO: add your codes here
x1, y1 = point1
x2, y2 = point2
print("x1:{}, y1:{}".format(x1, y1))
print("x2:{}, y2:{}".format(x2, y2))
center_x = (x1 + x2)/2
center_y = (y1 + y2)/2
print("center x: {}, y:{}".format(center_x, center_y))
center = (center_x, center_y)
return center
def line_center2(point1, point2):
return ((point1[0]+point2[0])/2, (point1[1] + point2[1])/2)
C = line_center(A, B) # C is center of line AB
mark(C, info="C", color="blue", size=10)
D = line_center2(A, B)
print(D)
mark(D, info="D", color="black", size=10)
###Output
_____no_output_____
###Markdown
4. find tow points on a line that divide the line into 3 equal parts.
###Code
from turtle import setup, reset, pu, pd, bye, left, right, fd, bk, screensize
from turtle import goto, seth, write, ht, st, home, dot, pen, speed
from qianglib import prepare_paper, draw_grid, mark, lines, line, polygon, text
width, height = 800, 600
setup(width, height, 0, 0)
prepare_paper(width, height, scale=20, min_x=0, min_y=0, max_y=25)
A = (10, 12)
B = (25, 21)
mark(A, info="A(10, 12)")
mark(B, info="B(25, 21)")
line(A, B)
def three_equal_division(point1, point2):
x1, y1 = point1
x2, y2 = point2
x_step = (x2 - x1)/3
y_step = (y2 - y1)/3
x3, y3 = x1 + x_step, y1 + y_step
x4, y4 = x2 - x_step, y2 - y_step
point3 = (x3, y3)
point4 = (x4, y4)
return [point3, point4]
result_list = three_equal_division(A, B)
E = result_list[0]
F = result_list[1]
mark(E)
mark(F)
line(A, E, color="yellow", linewidth=5)
line(E, F, color="red", linewidth=5)
line(F, B, color="blue", linewidth=5)
# Example 2
P1 = (10, 5)
P2 = (25, 5)
mark(P1)
mark(P2)
result_list = three_equal_division(P1, P2)
P3, P4 = result_list[0], result_list[1]
line(P1, P3, color="yellow")
line(P3, P4, color="red")
line(P4, P2, color="blue")
mark(P3, "P3(15, 5)")
mark(P4, "P4(20, 5)")
P5 = (5, 12)
P6 = (5, 21)
result_list = three_equal_division(P5, P6)
P7, P8 = result_list[0], result_list[1]
line(P5, P7, color="yellow")
line(P7, P8, color="red")
line(P8, P6, color="blue")
mark(P5)
mark(P6)
mark(P7)
mark(P8)
###Output
_____no_output_____ |
notebooks/LA_Assignment5.ipynb | ###Markdown
Assignment 5 - (Un)supervised machine learningApplying (un)supervised machine learning to text dataEITHERTrain a text classifier on your data to predict some label found in the metadata. For example, maybe you want to use this data to see if you can predict sentiment label based on text content.ORTrain an LDA model on your data to extract structured information that can provide insight into your data. For example, maybe you are interested in seeing how different authors cluster together or how concepts change over time in this dataset.You should formulate a short research statement explaining why you have chosen this dataset and what you hope to investigate. This only needs to be a paragraph or two long and should be included as a README file along with the code. E.g.: I chose this dataset because I am interested in... I wanted to see if it was possible to predict X for this corpus.In this case, your peer reviewer will not just be looking to the quality of your code. Instead, they'll also consider the whole project including choice of data, methods, and output. Think about how you want your output to look. Should there be visualizations? CSVs?You should also include a couple of paragraphs in the README on the results, so that a reader can make sense of it all. E.g.: I wanted to study if it was possible to predict X. The most successful model I trained had a weighted accuracy of 0.6, implying that it is not possible to predict X from the text content alone. And so on.TipsThink carefully about the kind of preprocessing steps your text data may require - and document these decisions!Your choice of data will (or should) dictate the task you choose - that is to say, some data are clearly more suited to supervised than unsupervised learning and vice versa. Make sure you use an appropriate method for the data and for the question you want to answerYour peer reviewer needs to see how you came to your results - they don't strictly speaking need lots of fancy command line arguments set up using argparse(). You should still try to have well-structured code, of course, but you can focus less on having a fully-featured command line toolBonus challengesDo both tasks - either with the same or different datasetsGeneral instructionsYou should upload standalone .py script(s) which can be executed from the command lineYou must include a requirements.txt file and a bash script to set up a virtual environment for the project You can use those on worker02 as a templateYou can either upload the scripts here or push to GitHub and include a link - or both!Your code should be clearly documented in a way that allows others to easily follow the structure of your script and to use them from the command linePurposeThis assignment is designed to test that you have an understanding of:how to formulate research projects with computational elements;how to perform (un)supervised machine learning on text data;how to present results in an accessible manner.
###Code
# standard library
import sys,os
sys.path.append(os.path.join(".."))
from pprint import pprint
from tqdm import tqdm
# data and nlp
import pandas as pd
import spacy
nlp = spacy.load("en_core_web_sm", disable=["ner"])
#
import nltk
from nltk.corpus import stopwords
stop_words = stopwords.words('english')
# visualisation
import pyLDAvis.gensim
pyLDAvis.enable_notebook()
import seaborn as sns
from matplotlib import rcParams
# figure size in inches
rcParams['figure.figsize'] = 20,10
# LDA tools
import gensim
import gensim.corpora as corpora
from gensim.models import CoherenceModel
from gensim.utils import simple_preprocess
from utils import lda_utils
# warnings
import logging, warnings
warnings.filterwarnings('ignore')
logging.basicConfig(format='%(asctime)s : %(levelname)s : %(message)s', level=logging.ERROR)
data = pd.read_csv(os.path.join("..", "data", "phil_nlp (5).csv")).sample(50000)
data.sample(10)
bigram = gensim.models.Phrases(data["sentence_str"], min_count=5, threshold=100) # higher threshold fewer phrases.
trigram = gensim.models.Phrases(bigram[data["sentence_str"]],min_count=3, threshold=100)
bigram_mod = gensim.models.phrases.Phraser(bigram)
trigram_mod = gensim.models.phrases.Phraser(trigram)
def process_words(texts, nlp, bigram_mod, trigram_mod, stop_words=stop_words, allowed_postags=['NOUN', "ADJ", "VERB", "ADV"]):
"""Remove Stopwords, Form Bigrams, Trigrams and Lemmatization"""
# use gensim simple preprocess
texts = [[word for word in simple_preprocess(str(doc)) if word not in stop_words] for doc in tqdm(texts)]
print("the texts have been roughly preprocessed")
texts = [bigram_mod[doc] for doc in tqdm(texts)]
print("bigrams have been formed")
texts = [trigram_mod[bigram_mod[doc]] for doc in tqdm(texts)]
print("trigrams have been formed")
texts_out = []
# lemmatize and POS tag using spaCy
print("lemmatizing and pos-tagging docs...")
for sent in tqdm(texts):
doc = nlp(" ".join(sent))
texts_out.append([f"{token.lemma_}_{token.pos_}" for token in doc if token.pos_ in allowed_postags])
return texts_out
texts_out = []
data_processed = process_words(data["sentence_str"],nlp, bigram_mod, trigram_mod, allowed_postags=["ADJ", "ADV", "NOUN", "PROPN", "VERB", "NUM"])
df =pd.DataFrame(data_processed)
data_processed[1]
pd.DataFrame(data_processed).to_csv(os.path.join("..", "data", "assignment5", "data_preprocessed.csv"))
with open(os.path.join("..", "data", "assignment5", "data_preprocessed.txt"), 'w') as file:
for item in data_processed:
file.write(" ".join(map(str,item)))
file.write("\n")
# read file in a string list
with open(os.path.join("..", "data", "assignment5", "data_preprocessed.txt")) as f:
lineList = f.readlines()
test = [[word for word in simple_preprocess(str(doc))] for doc in tqdm(lineList)]
test[0]
test = open(os.path.join("..", "data", "assignment5", "data_preprocessed.txt"), 'r', sep = "\n")
# Create Dictionary
id2word = corpora.Dictionary(data_processed)
# Create Corpus: Term Document Frequency
corpus = [id2word.doc2bow(text) for text in data_processed]
lda_model = gensim.models.LdaMulticore(corpus=corpus, # vectorised corpus - list of lists of tupols
id2word=id2word, # gensim dictionary - mapping words to IDS
num_topics=32, # number of topics
random_state=100, # set for reproducability
chunksize=10, # batch data for efficiency
passes=10, # number of full passess over data
iterations=100, # related to document rather than corpus
per_word_topics=True, # define word distributions
minimum_probability=0.0)
# Compute Perplexity
print('\nPerplexity: ', lda_model.log_perplexity(corpus)) # a measure of how good the model is. lower the better.
# Compute Coherence Score
coherence_model_lda = CoherenceModel(model=lda_model,
texts=data_processed,
dictionary=id2word,
coherence='c_v')
coherence_lda = coherence_model_lda.get_coherence()
print('\nCoherence Score: ', coherence_lda)
# Can take a long time to run.
model_list, coherence_values = lda_utils.compute_coherence_values(texts=data_processed,
corpus=corpus,
dictionary=id2word,
start=5,
limit=40,
step=5)
df_topic_keywords = lda_utils.format_topics_sentences(ldamodel=lda_model,
corpus=corpus,
texts=data_processed)
# Format
df_dominant_topic = df_topic_keywords.reset_index()
df_dominant_topic.columns = ['Chunk_No', 'Dominant_Topic', 'Topic_Perc_Contrib', 'Keywords', 'Text']
df_dominant_topic.sample(10)
# Display setting to show more characters in column
pd.options.display.max_colwidth = 100
sent_topics_sorteddf = pd.DataFrame()
sent_topics_outdf_grpd = df_topic_keywords.groupby('Dominant_Topic')
for i, grp in sent_topics_outdf_grpd:
sent_topics_sorteddf = pd.concat([sent_topics_sorteddf,
grp.sort_values(['Perc_Contribution'], ascending=False).head(1)],
axis=0)
# Reset Index
sent_topics_sorteddf.reset_index(drop=True, inplace=True)
# Format
sent_topics_sorteddf.columns = ['Topic_Num', "Topic_Perc_Contrib", "Keywords", "Representative Text"]
# Show
sent_topics_sorteddf.head(10)
vis = pyLDAvis.gensim.prepare(lda_model, corpus, dictionary=lda_model.id2word)
vis
###Output
_____no_output_____ |
_notebooks/2021-06-23-manim_sine_curve_colab.ipynb | ###Markdown
Sine curve unit circle on Colab> "manim example"- toc: false- branch: master- hidden: true- categories: [manim, colab]
###Code
%%capture
# https://docs.manim.community/en/stable/installation/colab.html
!sudo apt update
!sudo apt install libcairo2-dev ffmpeg texlive texlive-latex-extra texlive-fonts-extra texlive-latex-recommended texlive-science tipa libpango1.0-dev
!pip install manim
!pip install IPython --upgrade
from manim import *
%%manim -v WARNING --disable_caching -qm SineCurveUnitCircle
# https://docs.manim.community/en/stable/examples.html#sinecurveunitcircle
class SineCurveUnitCircle(Scene):
# contributed by heejin_park, https://infograph.tistory.com/230
def construct(self):
self.show_axis()
self.show_circle()
self.move_dot_and_draw_curve()
self.wait()
def show_axis(self):
x_start = np.array([-6,0,0])
x_end = np.array([6,0,0])
y_start = np.array([-4,-2,0])
y_end = np.array([-4,2,0])
x_axis = Line(x_start, x_end)
y_axis = Line(y_start, y_end)
self.add(x_axis, y_axis)
self.add_x_labels()
self.origin_point = np.array([-4,0,0])
self.curve_start = np.array([-3,0,0])
def add_x_labels(self):
x_labels = [
MathTex("\pi"), MathTex("2 \pi"),
MathTex("3 \pi"), MathTex("4 \pi"),
]
for i in range(len(x_labels)):
x_labels[i].next_to(np.array([-1 + 2*i, 0, 0]), DOWN)
self.add(x_labels[i])
def show_circle(self):
circle = Circle(radius=1)
circle.move_to(self.origin_point)
self.add(circle)
self.circle = circle
def move_dot_and_draw_curve(self):
orbit = self.circle
origin_point = self.origin_point
dot = Dot(radius=0.08, color=YELLOW)
dot.move_to(orbit.point_from_proportion(0))
self.t_offset = 0
rate = 0.25
def go_around_circle(mob, dt):
self.t_offset += (dt * rate)
# print(self.t_offset)
mob.move_to(orbit.point_from_proportion(self.t_offset % 1))
def get_line_to_circle():
return Line(origin_point, dot.get_center(), color=BLUE)
def get_line_to_curve():
x = self.curve_start[0] + self.t_offset * 4
y = dot.get_center()[1]
return Line(dot.get_center(), np.array([x,y,0]), color=YELLOW_A, stroke_width=2 )
self.curve = VGroup()
self.curve.add(Line(self.curve_start,self.curve_start))
def get_curve():
last_line = self.curve[-1]
x = self.curve_start[0] + self.t_offset * 4
y = dot.get_center()[1]
new_line = Line(last_line.get_end(),np.array([x,y,0]), color=YELLOW_D)
self.curve.add(new_line)
return self.curve
dot.add_updater(go_around_circle)
origin_to_circle_line = always_redraw(get_line_to_circle)
dot_to_curve_line = always_redraw(get_line_to_curve)
sine_curve_line = always_redraw(get_curve)
self.add(dot)
self.add(orbit, origin_to_circle_line, dot_to_curve_line, sine_curve_line)
self.wait(8.5)
dot.remove_updater(go_around_circle)
###Output
|
K-Meansapplicationtoimagecompression.ipynb | ###Markdown
K-Means clustering used for image compressionIn the following we discuss a curious application of K-Means clustering to image compression.In this notebook [this notebook](https://github.com/andreaspts/ML_KMeans_Clustering_Analyses/blob/master/simpleKmeansclusterungoncardata.ipynb), we introduced this particular algorithm and exemplified it's power on a simple example. For a reminder on K-Means clustering, we refer to https://en.wikipedia.org/wiki/K-means_clustering. In a first step, we have to load the image we would like to compress. We remind ourselves that the information encoding the image is in fact represented by an array. We then play around with this structure to gain some intuition about the dimensions of this array. We then reshape it to bring it into the appropriate form needed for the K-Means processing. In a second step, we feed it into the KMeans function and demand that the image information given in terms of numbers in the (now reshaped) array is clustered into 20 clusters. The idea is that e.g. reddish colors represented by a range of numbers are clustered around a mean color value. This is done for 20 color groups. We report these 20 cluster centers and the cluster labels, where the latter tells us which pixel belongs to which of the 20 clusters.Then these labels per pixel are associated to the cluster centers to regain the now compressed image/array which is subsequently reshaped.Finally, the image is printed and saved.
###Code
#import necessary packages
from skimage import io, exposure
import imageio
import numpy as np
#load image
image = io.imread("./cat2.jpg")
io.imshow(image)
#retrieve dimensional info of the image/array
image.shape
#would like to drop the transparency values (alpha channel in png files) if available
#image[:, :, [0, 1, 2]]
image[:, :, [0, 1, 2]].shape
#some basic image manipulations to get an intuition for the dimensions
#render picutre brighter
image_without_alpha = image[:, :, :3]
image_brighter = image_without_alpha + 40
image_brighter[image_brighter < image_without_alpha] = 255
io.imshow(image_brighter)
#dimensions without alpha channel
image_without_alpha.shape
#bring image array into different form needed for KMeans algorithm
image_reshaped = image_without_alpha.reshape(-1,3)
image_without_alpha.reshape(-1,3).shape
#use KMeans clustering to reduce colors of an image to 20
from sklearn.cluster import KMeans
model = KMeans(n_clusters = 20, n_init = 1) #20 clusters only @ 1 attempt
model.fit(image_reshaped)
#print center points of the 20 clusters
print(model.cluster_centers_)
#print cluster label per pixel
print(model.labels_)
#length of the cluster labels equals (of course) that dimension of the array storing the image info
len(model.labels_)
#associate pixel labels to cluster centers: "giving restored image"
colors = model.cluster_centers_.astype("uint8") #to convert centers values into integers representing true image information
pixels = model.labels_
#save restored image
np.savez_compressed("./image.npz", pixels = pixels, colors = colors)
with np.load("./image.npz") as file:
pixels = file["pixels"]
colors = file["colors"]
pixels_transformed = []
for pixel in pixels:
pixels_transformed.append(colors[pixel])
#transform back into np array
pixels_transformed = np.array(pixels_transformed)
pixels_transformed.shape
#reshaping in order to show the compressed image
image_restored = pixels_transformed.reshape(878, 1600, 3)
io.imshow(image_restored)
#saving compressed (array as an) image
imageio.imwrite('cat2_transformed.jpg', image_restored)
###Output
_____no_output_____ |
working_notebook/TaylorDataPrep.ipynb | ###Markdown
Remuve duplicates*same album appears multiple times, but with different names, and it is actually a problem
###Code
#Deleting the Live, Genius, Demo and stuff
k=0
Not_good_words=['Live','Genius','Demo','folklore']
TODEL=[]
for alb in data.album.drop_duplicates().tolist():
sp=alb.split()
if len(sp)!=0:
for s in sp:
if s=='Live' or s=='Genius' or s=='Demo' or s=='folklore': # checkif the tile one of this target words
TODEL.append(k)
k=k+1
alb=np.array(data.album.drop_duplicates().tolist())
ALB=[]
for a in range(len(alb)):
if a not in TODEL:
ALB.append(alb[a])
LIST_ALB_TITLE=[]
for a in ALB:
sp=a.split()
LIST_ALB_TITLE.append(sp)
print("ALB:",ALB)
# Adding folklore
ALB=ALB+['folklore']
ALB=pd.DataFrame(ALB).drop_duplicates(keep='first')
ALB=ALB.drop([15,12])[0].tolist()
#Taking the final good index and the choruses of these ones.
IND=data[(data.album.isin(ALB)) & (data.lyric=='[Chorus]')].drop_duplicates(subset='song_title').index.tolist()
#Identifying the stop (e.g. '[Chorus]')
stop=[]
for l in data.lyric.tolist():
if l[0]=='[':
stop.append(1)
else:
stop.append(0)
data['S']=stop
#Move in the index range
CHORUSES=[]
for i in IND:
#Pick the chorus till the following line
new_d=data[i+1::]
ss=new_d.S.tolist()
for q in range(len(ss)):
if ss[q]==1:
#Stop when you see that the choruses is ended
break
new_d=data[i+1:i+1+q]
#BOOM! You have the chorus
CHORUSES.append(new_d.lyric.tolist())
print("Number of chorouses:",len(CHORUSES))
nr_verse=0
for c in CHORUSES:
nr_verse=nr_verse+len(c)
print("Number of verses:", nr_verse)
bulk_corpus = " "
for chorus in CHORUSES:
for vers in chorus:
bulk_corpus= bulk_corpus +"\n"+vers
print(bulk_corpus)
text_file = open("TaylorSwift/bulk_taylor.txt", "w")
n = text_file.write(bulk_corpus)
text_file.close()
###Output
_____no_output_____ |
2_4_overfitting_underfitting/2_2_Aufgabe - Overfitting - Underfitting.ipynb | ###Markdown
In diesem Tutorial beschäftigen Sie sich anhand eines Spielbeispiels mit den Problemen einer Überanpassung oder Unteranpassung der linearen bzw. logistischen Regression.In der begleitenden Python-File `utils.py` befinden sich Hilfsfunktionen zum Erstellen eines zufälligen Trainings- und Testdatensatzes mit einer Beobachtung und einer kontinuierlichen Zielvariablen. (2.2.1) Lineare Regression (o) &x1F4D7;** (a) ** &x1F4D7; Erstellen Sie per `utils.get_train_data()` einen Trainingsdatensatz mit Inputvariablen $\{x^{(i)} \; | \; i = 1, ..., N\}$ und Zielvariablen $\{y_T^{(i)}\; | \; i = 1, ..., N\}$ und führen Sie darauf eine lineare Regression aus.** (b) ** &x1F4D7; Treffen Sie eine Vorhersage der Zielvariablen, $\{\hat{y}^{(i)}\; | \; i = 1, ..., N\}$, für die Beobachtungen des Trainingsdatensatzes. Beurteilen Sie die Qualität der Vorhersage, indem Sie einmal den durchschnittlichen quadratischen und einmal den durchschnittlichen absoluten Fehler der Vorhersage berechnen:(i) Quadratischer Fehler: $ \frac{1}{N} \sum_{i=1}^N (\hat{y}^{(i)} - y_T^{(i)})^2$(ii) Absoluter Fehler: $ \frac{1}{N} \sum_{i=1}^N | \hat{y}^{(i)} - y_T^{(i)} | $*(Tipp: wenn der quadratische Fehler aus Ihrer Sicht keine Aussagekraft hat, verwenden Sie stattdessen den RMSE)*** (c) ** &x1F4D7; Visualisieren Sie das Ergebnis der Regression auf eine geeignete Weise.** (d) ** &x1F4D7; Erstellen Sie nun einen Testdatensatz per `utils.get_test_data()` und treffen Sie erneut eine Vorhersage der Zielvariablen mit dem in **b)** erstellten Modell. Berechnen Sie den durchschnittlichen quadratischen und absoluten Fehler der Vorhersage auf dem Testdatensatz. Interpretieren Sie das Ergebnis. *(Tipp: wenn der quadratische Fehler aus Ihrer Sicht keine Aussagekraft hat, verwenden Sie stattdessen den RMSE)*** (e) ** &x1F4D7; Wiederholen Sie die Aufgaben **b)** bis **c)** für ein quadratisches Modell (Nutzen Sie dafür zum Beispiel `from sklearn.preprocessing import PolynomialFeatures`.). Interpretieren Sie die Ergebnisse.
###Code
import utils
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import PolynomialFeatures
from sklearn.pipeline import make_pipeline
import numpy as np
###Output
_____no_output_____
###Markdown
(2.2.2) Zufällige Trainingsdaten ** (oo) ** &x1F4D9; Die Hilfsfunktion `utils.get_train_data()` erzeugt bei jedem Aufruf einen neuen, zufälligen, Datensatz während die Funktion `utils.get_test_data()` einen festen Testdatensatz erzeugt. In dieser Aufgabe untersuchen Sie, welchen Einfluss die Zufälligkeit des Trainingsdatensatzes auf die Qualität des Modells hat.** (a) ** &x1F4D7; Erstellen und visualisieren Sie exemplarisch zwei verschiedene Trainingsdatensätze.** (b) ** &x1F4D9; Wiederholen Sie die Aufgaben **1a)**, **1b)** und **1d)** für $10-20$ zufällig generierte Trainingsdatensätze. Entscheiden Sie sich dabei für eine der Fehlermetriken (zum Beispiel RMSE). Speichern Sie sich die Fehler für jede der $10-20$ Wiederholungen des Experiments.Berechnen Sie dann folgende Größe: Für jeden Trainingsdatensatz haben Sie ein separates Modell trainiert und evaluiert. Daraus resultiert jeweils ein Trainingsfehler und ein Testfehler. Berechnen Sie nun den durchschnittlichen Trainingsfehler und Testfehler und die Standardabweichung dieser Fehler über alle Trainingsdatensätze hinweg (*Hinweis: der Trainings- und Testfehler sind für sich genommen schon Durchschnittswerte - nämlich über die Datenpunkte hinweg. Hier aber ist gemeint, die Durchschnittswerte dieser Fehler für die Widerholungen des Experiments zu berechnen - in einem gewissen Sinne also Durchschnittswerte der Durchschnittswerte*)** (c) ** &x1F4D9; Visualisieren Sie die Ergebnisse aus **c)** indem Sie die $10-20$ verschiedenen linearen Modelle in einem einzigen Plot darstellen.** (d) ** &x1F4D9; Wiederholen Sie nun die vorherigen Aufgabenteile während Sie anstelle eines linearen Modells ein quadratisches Modell oder sogar ein Modell höheren Grades verwenden (siehe Aufgabe **1d)**).** (e) ** &x1F4D9; Interpretieren Sie Ihre Ergebnisse. (2.2.3) Bias-Variance-Tradeoff (ooo) &x1F4D8; In der vorherigen Aufgabe haben Sie eine Reihe von Modellen auf der Basis zufälliger Trainingsdaten erstellt und für jedes Modell den Testfehler berechnet. Daraufhin ließ sich der durchschnittliche Testfehler sowie die Varianz des Testfehlers schätzen. Sie haben das lineare Modell mit dem quadratischen Modell verglichen.Nun wollen wir die Komplexität des Modells systematisch erhöhen.Als Maß für die Komplexität des Modells nehmen wir den Grad der polynomischen Expansion an. Der Parameter `'degree'` kann von $1$ (lineares Modell) systematisch erhöht werden. Für jede Komplexitätsstufe lassen sich dann eine Reihe Modelle auf Basis zufälliger Trainingsdaten erstellen. Der Testdatensatz bleibt stets derselbe.Wiederholen Sie für jeden Grad (`degree`) der polynomischen Expansion die folgenden Schritte:*(i)* Trainieren Sie $10-20$ verschiedene Modelle jeweils auf einem zufällig generierten Trainingsdatensatz. Um die gewünschten Ergebnisse sichtbar zu machen, bietet es sich an, die Menge an Beobachtungen noch weiter zu reduzieren. Nutzen Sie dafür das Argument `n_samples` der Funktion `utils.get_train_data()`.*(ii)* Berechnen Sie die durchschnittliche Vorhersage zwischen diesen Modellen und plotten Sie diese etwa für $x \in [0, 10]$.*(iii)* Berechnen Sie die Standardabweichung zwischen den verschiedenen Vorhersagen und visualisieren Sie diese auf eine geeignete Weise für $x \in [0, 10]$.*(iv)* Benutzen Sie `utils.true_function` um die den Daten tatsächlich zu Grunde liegende Funktion zu plotten. Versuchen Sie, die Plots aus *(ii)*-*(iv)* für jeden Grad der polynomischen Expansion in einem einzigen Plot darzustellen. Interpretieren Sie ihre Ergebnisse. (2.2.4) Regularisierung (o) - (oo) &x1F4D7; Um das Risiko einer Überanpassung zu verhindern, kann die lineare/polynomiale Regression regularisiert werden. Dazu wird der Verlustfunktion ein zusätzlicher Regularisierungsterm hinzugefügt, der dafür sorgt, dass Koeffizienten kleiner Magnitude gegenüber Koeffizienten großer Magnitude bevorzugt werden.Scikit-Learn stellt die lineare Regression mit Regularisierung in den Klassen `Ridge`, `ElasticNet` und `Lasso` zur Verfügung.** (a) ** &x1F4D7; Beschäftigen Sie sich zunächst der Dokumentation aller drei Klassen. Was ist der wesentliche Unterschied zwischen den Klassen? Benutzen Sie im Folgenden nur die Klasse `Ridge` für eine lineare Regression mit L2-Regularisierung. Setzen Sie in jedem Fall `normalize=True` für alle weiteren Experimente.** (b) ** &x1F4D7; Wählen Sie ein Regressionsmodell mit einem mittleren Grad der polynomischen Expansion, etwa 6-8. Generieren Sie zunächst einen Trainingsdatensatz wie in den vorherigen Aufgaben und fitten Sie das Modell. Vergleichen Sie die Ergebnisse einer Regression mit `alpha=0.0`, `alpha=1.0` und `alpha=10.0`, indem Sie den Fit wie in den vorherigen Aufgaben auf eine geeignete Weise visualisieren und die Trainings- und Testfehler der Verfahren miteinander vergleichen. Interpretieren Sie.** (c) ** &x1F4D9; Varieren Sie nun den Hyperparameter `alpha` der Regression systematisch, z.B. logarithmisch: $\alpha = 0, 10^{-3}, 5 \cdot 10^{-3}, 10^{-2}, ..., 10$ (Tipp: `np.logspace`). Trainieren Sie nun für jeden Wert des Hyperparameters $20-50$ verschiedene Modelle auf jeweils zufällig generierten Trainingsdaten und berechnen Sie jedesmal den Trainingsfehler sowie den Testfehler. Plotten Sie dann den durchschnittlichen Trainings- sowie Testfehler (über die zufälligen Trainingsdatensätze hinweg) sowie, in einem separaten Plot, deren Standardabweichung, gegen den Wert des Hyperparameters. Um das Ergebnis sichtbar zu machen, können Sie die Menge an Beobachtungen für die Trainingsdaten reduzieren, indem Sie das Argument `n_samples` der Funktion `utils.get_train_data()` verwenden. Interpretieren Sie das Ergebnis.
###Code
from sklearn.linear_model import Ridge, ElasticNet, Lasso
###Output
_____no_output_____ |
notebooks/data-owner_GTEx-regression.ipynb | ###Markdown
Federated Learning - GTEx_V8 Example Import dependencies
###Code
#dependencies for helper functions/classes
import pandas as pd
import pyarrow.parquet as pq
from typing import NamedTuple
import os.path as path
import os
import progressbar
import requests
import numpy as np
import random
#keras for ML
import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.layers import Dropout, Input, Dense
from tensorflow.keras.models import Sequential, load_model, Model
from tensorflow.keras.utils import plot_model, normalize
from tensorflow.keras import regularizers
from tensorflow.keras.optimizers import SGD, Adam, Nadam, Adadelta
from tensorflow.keras.activations import relu, elu, sigmoid
#sklearn for preprocessing the data and train-test split
from sklearn.utils import class_weight
from sklearn.model_selection import train_test_split, cross_val_score, KFold
from sklearn.preprocessing import OneHotEncoder, MinMaxScaler, LabelEncoder, StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.metrics import r2_score, mean_squared_error, mean_absolute_error, accuracy_score, classification_report
from sklearn.metrics import f1_score, accuracy_score, precision_score, recall_score, r2_score, mean_squared_error, mean_absolute_error
#for plots
import matplotlib
import matplotlib.pyplot as plt
#%matplotlib inline
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Conv2D
from tensorflow.keras.layers import MaxPooling2D
from tensorflow.keras.layers import Activation
from tensorflow.keras.layers import Flatten
from tensorflow.keras.layers import Dense
from tensorflow.keras.optimizers import SGD
from tensorflow.keras import backend as K
###Output
_____no_output_____
###Markdown
Parameter cell -->
###Code
seed = 7
_test_size = 0.2
comms_round = 50
local_epochs = 100
CLIENTS = 2
local_batch_size = 256
VERBOSE = 1
class Labels(NamedTuple):
'''
One-hot labeled data
'''
tissue: np.ndarray
sex: np.ndarray
age: np.ndarray
death: np.ndarray
class Genes:
'''
Class to load GTEX samples and gene expressions data
'''
def __init__(self, samples_path: str = '', expressions_path: str = '', problem_type: str = "classification"):
self.__set_samples(samples_path)
self.__set_labels(problem_type)
if expressions_path != '':
self.expressions = self.get_expressions(expressions_path)
def __set_samples(self, sample_path: str) -> pd.DataFrame:
self.samples: pd.DataFrame = pq.read_table(sample_path).to_pandas()
self.samples["Death"].fillna(-1.0, inplace = True)
self.samples: pd.DataFrame = self.samples.set_index("Name")
self.samples["Sex"].replace([1, 2], ['male', 'female'], inplace=True)
self.samples["Death"].replace([-1,0,1,2,3,4], ['alive/NA', 'ventilator case', '<10 min.', '<1 hr', '1-24 hr.', '>1 day'], inplace=True)
self.samples = self.samples[~self.samples['Death'].isin(['>1 day'])]
return self.samples
def __set_labels(self, problem_type: str = "classification") -> Labels:
self.labels_list = ["Tissue", "Sex", "Age", "Death"]
self.labels: pd.DataFrame = self.samples[self.labels_list]
self.drop_list = self.labels_list + ["Subtissue", "Avg_age"]
if problem_type == "classification":
dummies_df = pd.get_dummies(self.labels["Age"])
print(dummies_df.columns.tolist())
self.Y = dummies_df.values
if problem_type == "regression":
self.Y = self.samples["Avg_age"].values
return self.Y
def sex_output(self, model):
return Dense(units=self.Y.sex.shape[1], activation='softmax', name='sex_output')(model)
def tissue_output(self, model):
return Dense(units=self.Y.tissue.shape[1], activation='softmax', name='tissue_output')(model)
def death_output(self, model):
return Dense(units=self.Y.death.shape[1], activation='softmax', name='death_output')(model)
def age_output(self, model):
'''
Created an output layer for the keras mode
:param model: keras model
:return: keras Dense layer
'''
return Dense(units=self.Y.age.shape[1], activation='softmax', name='age_output')(model)
def get_expressions(self, expressions_path: str)->pd.DataFrame:
'''
load gene expressions DataFrame
:param expressions_path: path to file with expressions
:return: pandas dataframe with expression
'''
if expressions_path.endswith(".parquet"):
return pq.read_table(expressions_path).to_pandas().set_index("Name")
else:
separator = "," if expressions_path.endswith(".csv") else "\t"
return pd.read_csv(expressions_path, sep=separator).set_index("Name")
def prepare_data(self, normalize_expressions: bool = True)-> np.ndarray:
'''
:param normalize_expressions: if keras should normalize gene expressions
:return: X array to be used as input data by keras
'''
data = self.samples.join(self.expressions, on = "Name", how="inner")
ji = data.columns.drop(self.drop_list)
x = data[ji]
# adding one-hot-encoded tissues and sex
#x = pd.concat([x,pd.get_dummies(data['Tissue'], prefix='tissue'), pd.get_dummies(data['Sex'], prefix='sex')],axis=1)
steps = [('standardization', StandardScaler()), ('normalization', MinMaxScaler())]
pre_processing_pipeline = Pipeline(steps)
transformed_data = pre_processing_pipeline.fit_transform(x)
x = transformed_data
print('Data length', len(x))
return x #normalize(x, axis=0) if normalize_expressions else x
def get_features_dataframe(self, add_tissues=False):
data = self.samples.join(self.expressions, on = "Name", how="inner")
ji = data.columns.drop(self.drop_list)
df = data[ji]
if add_tissues:
df = pd.concat([df,pd.get_dummies(data['Tissue'], prefix='tissue'), pd.get_dummies(data['Sex'], prefix='sex')],axis=1)
x = df.values
min_max_scaler = MinMaxScaler()
x_scaled = min_max_scaler.fit_transform(x)
df_normalized = pd.DataFrame(x_scaled, columns=df.columns, index=df.index)
return df_normalized
samples_path = '../data/gtex/v8_samples.parquet'
expressions_path = '../data/gtex/v8_expressions.parquet'
genes = Genes(samples_path, expressions_path, problem_type="regression")
X = genes.prepare_data(True)
Y = genes.Y
#split data into training and test set
X_train, X_test, y_train, y_test = train_test_split(X,
Y,
test_size=_test_size,
random_state=seed)
def create_clients(image_list, label_list, num_clients=10, initial='clients'):
''' return: a dictionary with keys clients' names and value as
data shards - tuple of images and label lists.
args:
image_list: a list of numpy arrays of training images
label_list:a list of binarized labels for each image
num_client: number of fedrated members (clients)
initials: the clients'name prefix, e.g, clients_1
'''
#create a list of client names
client_names = ['{}_{}'.format(initial, i+1) for i in range(num_clients)]
#randomize the data
data = list(zip(image_list, label_list))
random.shuffle(data)
#shard data and place at each client
size = len(data)//num_clients
shards = [data[i:i + size] for i in range(0, size*num_clients, size)]
#number of clients must equal number of shards
assert(len(shards) == len(client_names))
return {client_names[i] : shards[i] for i in range(len(client_names))}
#create clients
clients = create_clients(X_train, y_train, num_clients=CLIENTS, initial='client')
clients.keys(), clients['client_1'][0][0].shape
def batch_data(data_shard, bs=256):
'''Takes in a clients data shard and create a tfds object off it
args:
shard: a data, label constituting a client's data shard
bs:batch size
return:
tfds object'''
#seperate shard into data and labels lists
data, label = zip(*data_shard)
dataset = tf.data.Dataset.from_tensor_slices((list(data), list(label)))
return dataset.shuffle(len(label)).batch(bs)
#process and batch the training data for each client
clients_batched = dict()
for (client_name, data) in clients.items():
clients_batched[client_name] = batch_data(data, bs = local_batch_size)
#process and batch the test set
test_batched = tf.data.Dataset.from_tensor_slices((X_test, y_test)).batch(len(y_test))
clients_batched.keys(),clients_batched['client_1']
from keras import backend as K
def coeff_determination(y_true, y_pred):
SS_res = K.sum(K.square( y_true-y_pred ))
SS_tot = K.sum(K.square( y_true - K.mean(y_true) ) )
return ( 1 - SS_res/(SS_tot + K.epsilon()) )
def optimized_age_model_regression():
# optimized_age_model(x_train, x_val, y_train, y_val, params: dict):
input_layer = Input(shape=(clients['client_1'][0][0].shape[0],))
reg = keras.regularizers.l1_l2(l1=0.3, l2=0.3)
mod = Dense(1024, activation=relu)(input_layer) # 196
mod = Dropout(0.1)(mod)
mod = Dense(512, activation=relu)(mod) # 196
mod = Dropout(0.1)(mod)
mod = Dense(64, activation=relu)(mod) #64
mod = Dropout(0.1)(mod)
outputs = [Dense(1, name='age_output')(mod)] #let's try to make it simple and start with age
#outputs = [Dense(y_train.shape[1], activation='sigmoid', name='age_output')(mod)] #let's try to make it simple and start with age
loss = {'age_output': 'mse'}
weights={'age_output': 1.0}
metrics = {'age_output': ['mae', coeff_determination]}
model = Model(inputs=input_layer, outputs=outputs)
model.summary()
model.compile(optimizer='adam',
loss=loss,
loss_weights=weights,
metrics=metrics,
)
return model
# class SimpleRegression:
# @staticmethod
# def build(shape = clients['client_1'][0][0].shape[0]):
# model = Sequential()
# model.add(Dense(1024, input_shape=(shape,)))
# model.add(Activation("relu"))
# model.add(Dropout(0.1))
# model.add(Dense(512))
# model.add(Activation("relu"))
# model.add(Dropout(0.1))
# model.add(Dense(64))
# model.add(Activation("relu"))
# model.add(Dropout(0.1))
# model.add(Dense(1))
# return model
# def global_model_init():
# model = Sequential()
# model.add(Dense(1024, input_shape=(clients['client_1'][0][0].shape[0],)))
# model.add(Activation("relu"))
# model.add(Dropout(0.1))
# model.add(Dense(512))
# model.add(Activation("relu"))
# model.add(Dropout(0.1))
# model.add(Dense(64))
# model.add(Activation("relu"))
# model.add(Dropout(0.1))
# model.add(Dense(1))
# return model
def Huber(yHat, y, delta=1.):
return np.where(np.abs(y-yHat) < delta,.5*(y-yHat)**2 , delta*(np.abs(y-yHat)-0.5*delta))
def transform_to_probas(age_intervals):
class_names = ['20-29', '30-39', '40-49', '50-59', '60-69', '70-79']
res = []
for a in age_intervals:
non_zero_index = class_names.index(a)
res.append([0 if i != non_zero_index else 1 for i in range(len(class_names))])
return np.array(res)
def transform_to_interval(age_probas):
class_names = ['20-29', '30-39', '40-49', '50-59', '60-69', '70-79']
return np.array(list(map(lambda p: class_names[np.argmax(p)], age_probas)))
def weight_scalling_factor(clients_trn_data, client_name):
client_names = list(clients_trn_data.keys())
#get the bs
bs = list(clients_trn_data[client_name])[0][0].shape[0]
#first calculate the total training data points across clinets
global_count = sum([tf.data.experimental.cardinality(clients_trn_data[client_name]).numpy() for client_name in client_names])*bs
# get the total number of data points held by a client
local_count = tf.data.experimental.cardinality(clients_trn_data[client_name]).numpy()*bs
return local_count/global_count
def scale_model_weights(weight, scalar):
'''function for scaling a models weights'''
weight_final = []
steps = len(weight)
for i in range(steps):
weight_final.append(scalar * weight[i])
return weight_final
def sum_scaled_weights(scaled_weight_list):
'''Return the sum of the listed scaled weights. The is equivalent to scaled avg of the weights'''
avg_grad = list()
#get the average grad accross all client gradients
for grad_list_tuple in zip(*scaled_weight_list):
layer_mean = tf.math.reduce_sum(grad_list_tuple, axis=0)
avg_grad.append(layer_mean)
return avg_grad
rmse = []
mae = []
r2 = []
huber_loss = []
# loss = 'mse'
# metrics = ['mae', coeff_determination]
# loss = {'age_output': 'mse'}
# weights={'age_output': 1.0}
# metrics = {'age_output': ['mae', coeff_determination]}
#initialize global model
global_model = optimized_age_model_regression()
# smlp_global = SimpleRegression()
# global_model = smlp_global.build()
#commence global training loop
for comm_round in range(comms_round):
print('='*62)
print('---------<STARTING TRAINING FOR ROUND {}>-----------'.format(comm_round))
# get the global model's weights - will serve as the initial weights for all local models
global_weights = global_model.get_weights()
#initial list to collect local model weights after scalling
scaled_local_weight_list = list()
#randomize client data - using keys
client_names= list(clients_batched.keys())
random.shuffle(client_names)
#loop through each client and create new local model
for client in client_names:
print('---------<STARTING TRAINING FOR CLIENT {}>-----------'.format(client))
local_model = optimized_age_model_regression()
# smlp_local = SimpleRegression()
# local_model = smlp_local.build()
# local_model.compile(loss=loss,
# optimizer='adam',
# metrics=metrics)
#set local model weight to the weight of the global model
local_model.set_weights(global_weights)
local_model.fit(clients_batched[client], epochs=local_epochs, verbose=VERBOSE)
predictions = local_model.predict(X_test)
test_y = y_test
print('---------<TEST RESULTS FOR CLIENT {} ; USING LOCAL MODEL>-----------'.format(client))
print("R^2", r2_score(test_y, predictions))
print("Mean squared error", mean_squared_error(test_y, predictions))
print("Mean absolute error", mean_absolute_error(test_y, predictions))
print('Huber loss', np.mean(Huber(test_y, predictions)))
#scale the model weights and add to list
scaling_factor = weight_scalling_factor(clients_batched, client)
print("SCALING FACTOR : {0}".format(scaling_factor))
scaled_weights = scale_model_weights(local_model.get_weights(), scaling_factor)
scaled_local_weight_list.append(scaled_weights)
#clear session to free memory after each communication round
K.clear_session()
#to get the average over all the local model, we simply take the sum of the scaled weights
average_weights = sum_scaled_weights(scaled_local_weight_list)
#update global model
global_model.set_weights(average_weights)
predictions = global_model.predict(X_test)
test_y = y_test
print('--------<TEST RESULTS AFTER ROUND {} ; USING GLOBAL MODEL>---------'.format(comm_round))
print("R^2", round(r2_score(test_y, predictions), 3))
print("Mean squared error", round(mean_squared_error(test_y, predictions), 3))
print("Mean absolute error", round(mean_absolute_error(test_y, predictions), 3))
print('Huber loss', round(np.mean(Huber(test_y, predictions)), 3))
rmse.append(mean_squared_error(test_y, predictions))
mae.append(mean_absolute_error(test_y, predictions))
r2.append(r2_score(test_y, predictions))
huber_loss.append(np.mean(Huber(test_y, predictions)))
print('='*62)
###Output
Model: "model_4"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_5 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense_12 (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout_12 (Dropout) (None, 1024) 0
_________________________________________________________________
dense_13 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_13 (Dropout) (None, 512) 0
_________________________________________________________________
dense_14 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_14 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
==============================================================
---------<STARTING TRAINING FOR ROUND 0>-----------
---------<STARTING TRAINING FOR CLIENT client_2>-----------
Model: "model_5"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_6 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense_15 (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout_15 (Dropout) (None, 1024) 0
_________________________________________________________________
dense_16 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_16 (Dropout) (None, 512) 0
_________________________________________________________________
dense_17 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_17 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 22s 900ms/step - loss: 711.8218 - mae: 21.4210 - coeff_determination: -3.4491
Epoch 2/100
24/24 [==============================] - 14s 578ms/step - loss: 330.9977 - mae: 14.4150 - coeff_determination: -1.0234
Epoch 3/100
24/24 [==============================] - 14s 585ms/step - loss: 276.4013 - mae: 13.2223 - coeff_determination: -0.6958
Epoch 4/100
24/24 [==============================] - 14s 589ms/step - loss: 235.7928 - mae: 12.2932 - coeff_determination: -0.4533
Epoch 5/100
24/24 [==============================] - 15s 633ms/step - loss: 214.9347 - mae: 11.7453 - coeff_determination: -0.3183
Epoch 6/100
24/24 [==============================] - 18s 736ms/step - loss: 197.0365 - mae: 11.2304 - coeff_determination: -0.2024
Epoch 7/100
24/24 [==============================] - 24s 1s/step - loss: 176.7989 - mae: 10.6877 - coeff_determination: -0.0827
Epoch 8/100
24/24 [==============================] - 16s 659ms/step - loss: 167.7708 - mae: 10.4220 - coeff_determination: -0.0316
Epoch 9/100
24/24 [==============================] - 14s 573ms/step - loss: 158.8795 - mae: 10.1070 - coeff_determination: 0.0283
Epoch 10/100
24/24 [==============================] - 13s 557ms/step - loss: 161.9794 - mae: 10.2786 - coeff_determination: 0.0089
Epoch 11/100
24/24 [==============================] - 14s 567ms/step - loss: 155.8361 - mae: 10.0048 - coeff_determination: 0.0434
Epoch 12/100
24/24 [==============================] - 13s 553ms/step - loss: 133.2294 - mae: 9.2436 - coeff_determination: 0.1842
Epoch 13/100
24/24 [==============================] - 13s 556ms/step - loss: 128.0723 - mae: 9.0768 - coeff_determination: 0.2196
Epoch 14/100
24/24 [==============================] - 13s 557ms/step - loss: 130.5900 - mae: 9.2220 - coeff_determination: 0.1978
Epoch 15/100
24/24 [==============================] - 13s 551ms/step - loss: 117.0472 - mae: 8.6489 - coeff_determination: 0.2818
Epoch 16/100
24/24 [==============================] - 14s 567ms/step - loss: 130.2188 - mae: 9.1692 - coeff_determination: 0.2010
Epoch 17/100
24/24 [==============================] - 13s 556ms/step - loss: 113.5172 - mae: 8.5855 - coeff_determination: 0.2932
Epoch 18/100
24/24 [==============================] - 13s 551ms/step - loss: 112.0882 - mae: 8.4862 - coeff_determination: 0.3082
Epoch 19/100
24/24 [==============================] - 13s 560ms/step - loss: 103.5279 - mae: 8.1372 - coeff_determination: 0.3624
Epoch 20/100
24/24 [==============================] - 14s 577ms/step - loss: 114.1549 - mae: 8.6251 - coeff_determination: 0.2850
Epoch 21/100
24/24 [==============================] - 14s 577ms/step - loss: 101.5476 - mae: 8.0213 - coeff_determination: 0.3835
Epoch 22/100
24/24 [==============================] - 14s 570ms/step - loss: 105.2950 - mae: 8.2293 - coeff_determination: 0.3547
Epoch 23/100
24/24 [==============================] - 13s 562ms/step - loss: 100.1649 - mae: 7.9865 - coeff_determination: 0.3783
Epoch 24/100
24/24 [==============================] - 13s 559ms/step - loss: 104.9791 - mae: 8.1861 - coeff_determination: 0.3567
Epoch 25/100
24/24 [==============================] - 14s 568ms/step - loss: 95.4008 - mae: 7.8091 - coeff_determination: 0.4128
Epoch 26/100
24/24 [==============================] - 13s 560ms/step - loss: 111.8253 - mae: 8.4757 - coeff_determination: 0.3233
Epoch 27/100
24/24 [==============================] - 14s 563ms/step - loss: 91.1053 - mae: 7.5975 - coeff_determination: 0.4379
Epoch 28/100
24/24 [==============================] - 14s 569ms/step - loss: 85.2423 - mae: 7.4103 - coeff_determination: 0.4723
Epoch 29/100
24/24 [==============================] - 13s 559ms/step - loss: 89.5538 - mae: 7.5504 - coeff_determination: 0.4520
Epoch 30/100
24/24 [==============================] - 14s 572ms/step - loss: 82.1167 - mae: 7.1936 - coeff_determination: 0.4989
Epoch 31/100
24/24 [==============================] - 15s 613ms/step - loss: 78.5589 - mae: 7.0683 - coeff_determination: 0.5173
Epoch 32/100
24/24 [==============================] - 13s 562ms/step - loss: 86.0092 - mae: 7.4139 - coeff_determination: 0.4701
Epoch 33/100
24/24 [==============================] - 13s 556ms/step - loss: 75.6967 - mae: 6.9227 - coeff_determination: 0.5358
Epoch 34/100
24/24 [==============================] - 14s 563ms/step - loss: 91.3069 - mae: 7.6531 - coeff_determination: 0.4386
Epoch 35/100
24/24 [==============================] - 13s 558ms/step - loss: 90.4786 - mae: 7.6718 - coeff_determination: 0.4424
Epoch 36/100
24/24 [==============================] - 13s 561ms/step - loss: 95.2333 - mae: 7.7981 - coeff_determination: 0.4196
Epoch 37/100
24/24 [==============================] - 13s 556ms/step - loss: 76.7806 - mae: 6.9583 - coeff_determination: 0.5318
Epoch 38/100
24/24 [==============================] - 13s 558ms/step - loss: 70.5146 - mae: 6.6443 - coeff_determination: 0.5680
Epoch 39/100
24/24 [==============================] - 14s 568ms/step - loss: 69.7687 - mae: 6.6130 - coeff_determination: 0.5737
Epoch 40/100
24/24 [==============================] - 13s 557ms/step - loss: 70.1452 - mae: 6.6319 - coeff_determination: 0.5696
Epoch 41/100
24/24 [==============================] - 14s 563ms/step - loss: 116.7408 - mae: 8.7470 - coeff_determination: 0.2746
Epoch 42/100
24/24 [==============================] - 13s 558ms/step - loss: 83.0586 - mae: 7.2809 - coeff_determination: 0.4836
Epoch 43/100
24/24 [==============================] - 13s 559ms/step - loss: 67.2713 - mae: 6.5226 - coeff_determination: 0.5830
Epoch 44/100
24/24 [==============================] - 14s 567ms/step - loss: 65.2352 - mae: 6.4024 - coeff_determination: 0.5995
Epoch 45/100
24/24 [==============================] - 13s 559ms/step - loss: 67.8142 - mae: 6.5167 - coeff_determination: 0.5811
Epoch 46/100
24/24 [==============================] - 13s 555ms/step - loss: 71.7981 - mae: 6.7153 - coeff_determination: 0.5610
Epoch 47/100
24/24 [==============================] - 13s 558ms/step - loss: 68.7351 - mae: 6.6475 - coeff_determination: 0.5664
Epoch 48/100
24/24 [==============================] - 14s 572ms/step - loss: 90.1364 - mae: 7.5820 - coeff_determination: 0.4523
Epoch 49/100
24/24 [==============================] - 14s 567ms/step - loss: 69.4793 - mae: 6.6106 - coeff_determination: 0.5750
Epoch 50/100
24/24 [==============================] - 14s 581ms/step - loss: 61.3960 - mae: 6.1966 - coeff_determination: 0.6206
Epoch 51/100
24/24 [==============================] - 14s 566ms/step - loss: 68.7231 - mae: 6.5611 - coeff_determination: 0.5783
Epoch 52/100
24/24 [==============================] - 13s 557ms/step - loss: 59.4956 - mae: 6.1222 - coeff_determination: 0.6344
Epoch 53/100
24/24 [==============================] - 14s 565ms/step - loss: 58.8928 - mae: 6.0806 - coeff_determination: 0.6385
Epoch 54/100
24/24 [==============================] - 13s 557ms/step - loss: 64.8401 - mae: 6.3740 - coeff_determination: 0.6043
Epoch 55/100
24/24 [==============================] - 13s 559ms/step - loss: 66.3404 - mae: 6.4920 - coeff_determination: 0.5918
Epoch 56/100
24/24 [==============================] - 13s 559ms/step - loss: 62.3591 - mae: 6.2478 - coeff_determination: 0.6179
Epoch 57/100
24/24 [==============================] - 13s 558ms/step - loss: 58.6166 - mae: 6.0465 - coeff_determination: 0.6408
Epoch 58/100
24/24 [==============================] - 14s 568ms/step - loss: 56.0955 - mae: 5.9443 - coeff_determination: 0.6548
Epoch 59/100
24/24 [==============================] - 13s 558ms/step - loss: 65.2508 - mae: 6.4151 - coeff_determination: 0.6031
Epoch 60/100
24/24 [==============================] - 13s 561ms/step - loss: 60.7987 - mae: 6.1865 - coeff_determination: 0.6244
Epoch 61/100
24/24 [==============================] - 13s 559ms/step - loss: 56.5023 - mae: 5.9721 - coeff_determination: 0.6500
Epoch 62/100
24/24 [==============================] - 14s 571ms/step - loss: 63.3025 - mae: 6.2890 - coeff_determination: 0.6123
Epoch 63/100
24/24 [==============================] - 13s 556ms/step - loss: 60.7650 - mae: 6.2591 - coeff_determination: 0.6153
Epoch 64/100
24/24 [==============================] - 13s 554ms/step - loss: 66.5197 - mae: 6.4951 - coeff_determination: 0.5896
Epoch 65/100
24/24 [==============================] - 14s 567ms/step - loss: 57.7542 - mae: 6.0692 - coeff_determination: 0.6421
Epoch 66/100
24/24 [==============================] - 13s 559ms/step - loss: 51.9207 - mae: 5.6892 - coeff_determination: 0.6790
Epoch 67/100
24/24 [==============================] - 14s 566ms/step - loss: 56.2244 - mae: 5.9350 - coeff_determination: 0.6562
Epoch 68/100
24/24 [==============================] - 13s 557ms/step - loss: 67.5080 - mae: 6.4815 - coeff_determination: 0.5858
Epoch 69/100
24/24 [==============================] - 13s 562ms/step - loss: 54.3223 - mae: 5.8546 - coeff_determination: 0.6656
Epoch 70/100
24/24 [==============================] - 13s 556ms/step - loss: 49.2176 - mae: 5.5483 - coeff_determination: 0.6968
Epoch 71/100
24/24 [==============================] - 14s 567ms/step - loss: 49.6234 - mae: 5.6467 - coeff_determination: 0.6865
Epoch 72/100
24/24 [==============================] - 14s 567ms/step - loss: 58.6433 - mae: 6.0596 - coeff_determination: 0.6442
Epoch 73/100
24/24 [==============================] - 13s 554ms/step - loss: 54.9162 - mae: 5.8795 - coeff_determination: 0.6648
Epoch 74/100
24/24 [==============================] - 13s 557ms/step - loss: 51.3651 - mae: 5.7289 - coeff_determination: 0.6791
Epoch 75/100
24/24 [==============================] - 13s 554ms/step - loss: 47.3033 - mae: 5.4892 - coeff_determination: 0.7049
Epoch 76/100
24/24 [==============================] - 14s 573ms/step - loss: 46.8704 - mae: 5.3867 - coeff_determination: 0.7111
Epoch 77/100
24/24 [==============================] - 13s 557ms/step - loss: 73.5117 - mae: 6.9758 - coeff_determination: 0.5412
Epoch 78/100
24/24 [==============================] - 13s 558ms/step - loss: 58.5556 - mae: 6.0767 - coeff_determination: 0.6415
Epoch 79/100
24/24 [==============================] - 13s 555ms/step - loss: 55.1858 - mae: 5.9063 - coeff_determination: 0.6619
Epoch 80/100
24/24 [==============================] - 14s 563ms/step - loss: 51.9686 - mae: 5.6948 - coeff_determination: 0.6816
Epoch 81/100
24/24 [==============================] - 14s 564ms/step - loss: 51.9255 - mae: 5.7185 - coeff_determination: 0.6833
Epoch 82/100
24/24 [==============================] - 13s 557ms/step - loss: 49.3878 - mae: 5.5627 - coeff_determination: 0.6950
Epoch 83/100
24/24 [==============================] - 13s 556ms/step - loss: 58.8007 - mae: 6.0525 - coeff_determination: 0.6469
Epoch 84/100
24/24 [==============================] - 13s 553ms/step - loss: 47.4369 - mae: 5.4238 - coeff_determination: 0.7108
Epoch 85/100
24/24 [==============================] - 13s 560ms/step - loss: 46.7652 - mae: 5.4130 - coeff_determination: 0.7081
Epoch 86/100
24/24 [==============================] - 14s 565ms/step - loss: 41.1738 - mae: 5.0609 - coeff_determination: 0.7461
Epoch 87/100
24/24 [==============================] - 13s 557ms/step - loss: 44.2818 - mae: 5.2721 - coeff_determination: 0.7254
Epoch 88/100
24/24 [==============================] - 13s 559ms/step - loss: 49.5031 - mae: 5.5655 - coeff_determination: 0.6980
Epoch 89/100
24/24 [==============================] - 13s 551ms/step - loss: 48.2639 - mae: 5.4574 - coeff_determination: 0.7042
Epoch 90/100
24/24 [==============================] - 13s 558ms/step - loss: 40.7611 - mae: 5.0543 - coeff_determination: 0.7473
Epoch 91/100
24/24 [==============================] - 14s 574ms/step - loss: 41.0846 - mae: 5.0488 - coeff_determination: 0.7472
Epoch 92/100
24/24 [==============================] - 13s 556ms/step - loss: 39.1620 - mae: 4.9758 - coeff_determination: 0.7588
Epoch 93/100
24/24 [==============================] - 14s 563ms/step - loss: 45.5177 - mae: 5.3946 - coeff_determination: 0.7178
Epoch 94/100
24/24 [==============================] - 13s 561ms/step - loss: 61.7304 - mae: 6.3221 - coeff_determination: 0.6193
Epoch 95/100
24/24 [==============================] - 14s 567ms/step - loss: 43.4926 - mae: 5.2049 - coeff_determination: 0.7328
Epoch 96/100
24/24 [==============================] - 13s 556ms/step - loss: 46.5659 - mae: 5.4080 - coeff_determination: 0.7120
Epoch 97/100
24/24 [==============================] - 14s 573ms/step - loss: 45.1703 - mae: 5.2954 - coeff_determination: 0.7261
Epoch 98/100
24/24 [==============================] - 13s 562ms/step - loss: 44.2005 - mae: 5.2717 - coeff_determination: 0.7241
Epoch 99/100
24/24 [==============================] - 13s 561ms/step - loss: 46.1151 - mae: 5.3460 - coeff_determination: 0.7166
Epoch 100/100
24/24 [==============================] - 14s 563ms/step - loss: 50.9457 - mae: 5.6311 - coeff_determination: 0.6888
---------<TEST RESULTS FOR CLIENT client_2 ; USING LOCAL MODEL>-----------
R^2 0.45038552601184345
Mean squared error 89.2555474094023
Mean absolute error 7.376016518004596
Huber loss 12.433490193488687
SCALING FACTOR : 0.5
---------<STARTING TRAINING FOR CLIENT client_1>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 18s 765ms/step - loss: 699.8377 - mae: 21.3000 - coeff_determination: -3.2201
Epoch 2/100
24/24 [==============================] - 13s 562ms/step - loss: 346.2864 - mae: 14.8558 - coeff_determination: -1.1155
Epoch 3/100
24/24 [==============================] - 13s 555ms/step - loss: 279.2757 - mae: 13.3528 - coeff_determination: -0.7223
Epoch 4/100
24/24 [==============================] - 14s 564ms/step - loss: 246.8430 - mae: 12.5409 - coeff_determination: -0.5188
Epoch 5/100
24/24 [==============================] - 13s 555ms/step - loss: 220.1531 - mae: 11.9045 - coeff_determination: -0.3486
Epoch 6/100
24/24 [==============================] - 13s 556ms/step - loss: 201.9059 - mae: 11.4694 - coeff_determination: -0.2517
Epoch 7/100
24/24 [==============================] - 13s 554ms/step - loss: 194.3162 - mae: 11.1995 - coeff_determination: -0.1953
Epoch 8/100
24/24 [==============================] - 13s 557ms/step - loss: 167.2331 - mae: 10.4263 - coeff_determination: -0.0286
Epoch 9/100
24/24 [==============================] - 14s 563ms/step - loss: 161.0272 - mae: 10.1876 - coeff_determination: 0.0113
Epoch 10/100
24/24 [==============================] - 13s 562ms/step - loss: 153.6948 - mae: 10.0337 - coeff_determination: 0.0460
Epoch 11/100
24/24 [==============================] - 14s 565ms/step - loss: 139.8444 - mae: 9.5116 - coeff_determination: 0.1410
Epoch 12/100
24/24 [==============================] - 13s 560ms/step - loss: 134.0726 - mae: 9.3932 - coeff_determination: 0.1589
Epoch 13/100
24/24 [==============================] - 14s 577ms/step - loss: 132.0951 - mae: 9.2048 - coeff_determination: 0.1909
Epoch 14/100
24/24 [==============================] - 13s 557ms/step - loss: 125.7268 - mae: 9.0600 - coeff_determination: 0.2169
Epoch 15/100
24/24 [==============================] - 14s 603ms/step - loss: 115.0094 - mae: 8.6540 - coeff_determination: 0.2834
Epoch 16/100
24/24 [==============================] - 13s 560ms/step - loss: 110.6596 - mae: 8.4345 - coeff_determination: 0.3171
Epoch 17/100
24/24 [==============================] - 13s 555ms/step - loss: 111.0611 - mae: 8.4476 - coeff_determination: 0.3171
Epoch 18/100
24/24 [==============================] - 14s 567ms/step - loss: 113.3457 - mae: 8.5543 - coeff_determination: 0.3015
Epoch 19/100
24/24 [==============================] - 14s 565ms/step - loss: 119.6151 - mae: 8.8404 - coeff_determination: 0.2595
Epoch 20/100
24/24 [==============================] - 14s 565ms/step - loss: 108.2603 - mae: 8.3687 - coeff_determination: 0.3322
Epoch 21/100
24/24 [==============================] - 13s 558ms/step - loss: 126.8991 - mae: 9.0666 - coeff_determination: 0.2178
Epoch 22/100
24/24 [==============================] - 13s 561ms/step - loss: 129.7868 - mae: 9.1379 - coeff_determination: 0.2063
Epoch 23/100
24/24 [==============================] - 13s 557ms/step - loss: 99.8798 - mae: 7.9689 - coeff_determination: 0.3890
Epoch 24/100
24/24 [==============================] - 13s 555ms/step - loss: 94.2041 - mae: 7.7857 - coeff_determination: 0.4222
Epoch 25/100
24/24 [==============================] - 13s 557ms/step - loss: 91.1295 - mae: 7.6208 - coeff_determination: 0.4399
Epoch 26/100
24/24 [==============================] - 14s 563ms/step - loss: 83.6288 - mae: 7.3027 - coeff_determination: 0.4839
Epoch 27/100
24/24 [==============================] - 13s 562ms/step - loss: 92.4648 - mae: 7.7800 - coeff_determination: 0.4200
Epoch 28/100
24/24 [==============================] - 13s 558ms/step - loss: 86.7600 - mae: 7.4108 - coeff_determination: 0.4694
Epoch 29/100
24/24 [==============================] - 13s 560ms/step - loss: 82.0953 - mae: 7.2449 - coeff_determination: 0.4905
Epoch 30/100
24/24 [==============================] - 13s 560ms/step - loss: 84.9045 - mae: 7.3336 - coeff_determination: 0.4831
Epoch 31/100
24/24 [==============================] - 13s 557ms/step - loss: 84.0527 - mae: 7.3686 - coeff_determination: 0.4817
Epoch 32/100
24/24 [==============================] - 14s 568ms/step - loss: 82.0975 - mae: 7.2692 - coeff_determination: 0.4921
Epoch 33/100
24/24 [==============================] - 13s 562ms/step - loss: 78.1414 - mae: 7.0197 - coeff_determination: 0.5188
Epoch 34/100
24/24 [==============================] - 13s 555ms/step - loss: 75.9149 - mae: 7.0073 - coeff_determination: 0.5337
Epoch 35/100
24/24 [==============================] - 13s 562ms/step - loss: 77.9862 - mae: 7.1193 - coeff_determination: 0.5137
Epoch 36/100
24/24 [==============================] - 14s 567ms/step - loss: 75.4096 - mae: 6.9360 - coeff_determination: 0.5393
Epoch 37/100
24/24 [==============================] - 14s 572ms/step - loss: 75.0134 - mae: 6.9164 - coeff_determination: 0.5389
Epoch 38/100
24/24 [==============================] - 14s 568ms/step - loss: 71.9818 - mae: 6.7784 - coeff_determination: 0.5565
Epoch 39/100
24/24 [==============================] - 13s 562ms/step - loss: 99.1934 - mae: 7.9884 - coeff_determination: 0.3932
Epoch 40/100
24/24 [==============================] - 13s 561ms/step - loss: 86.0637 - mae: 7.4720 - coeff_determination: 0.4706
Epoch 41/100
24/24 [==============================] - 14s 567ms/step - loss: 70.0507 - mae: 6.6958 - coeff_determination: 0.5657
Epoch 42/100
24/24 [==============================] - 13s 559ms/step - loss: 66.2854 - mae: 6.5676 - coeff_determination: 0.5821
Epoch 43/100
24/24 [==============================] - 14s 564ms/step - loss: 67.1188 - mae: 6.5566 - coeff_determination: 0.5862
Epoch 44/100
24/24 [==============================] - 13s 561ms/step - loss: 67.0474 - mae: 6.5448 - coeff_determination: 0.5838
Epoch 45/100
24/24 [==============================] - 13s 557ms/step - loss: 80.9741 - mae: 7.3465 - coeff_determination: 0.4784
Epoch 46/100
24/24 [==============================] - 14s 566ms/step - loss: 72.2880 - mae: 6.8070 - coeff_determination: 0.5535
Epoch 47/100
24/24 [==============================] - 13s 562ms/step - loss: 66.9639 - mae: 6.5849 - coeff_determination: 0.5855
Epoch 48/100
24/24 [==============================] - 13s 559ms/step - loss: 58.7379 - mae: 6.1181 - coeff_determination: 0.6383
Epoch 49/100
24/24 [==============================] - 13s 561ms/step - loss: 57.4229 - mae: 6.0369 - coeff_determination: 0.6477
Epoch 50/100
24/24 [==============================] - 13s 559ms/step - loss: 76.6898 - mae: 6.9965 - coeff_determination: 0.5297
Epoch 51/100
24/24 [==============================] - 14s 565ms/step - loss: 72.7064 - mae: 6.8072 - coeff_determination: 0.5565
Epoch 52/100
24/24 [==============================] - 13s 559ms/step - loss: 62.1031 - mae: 6.2586 - coeff_determination: 0.6178
Epoch 53/100
24/24 [==============================] - 14s 566ms/step - loss: 74.7076 - mae: 6.9352 - coeff_determination: 0.5354
Epoch 54/100
24/24 [==============================] - 13s 559ms/step - loss: 63.4820 - mae: 6.3159 - coeff_determination: 0.6083
Epoch 55/100
24/24 [==============================] - 14s 573ms/step - loss: 58.4122 - mae: 6.0567 - coeff_determination: 0.6386
Epoch 56/100
24/24 [==============================] - 14s 567ms/step - loss: 55.5519 - mae: 5.9556 - coeff_determination: 0.6560
Epoch 57/100
24/24 [==============================] - 13s 561ms/step - loss: 62.9743 - mae: 6.3226 - coeff_determination: 0.6133
Epoch 58/100
24/24 [==============================] - 13s 559ms/step - loss: 57.9091 - mae: 6.0516 - coeff_determination: 0.6442
Epoch 59/100
24/24 [==============================] - 14s 573ms/step - loss: 53.5871 - mae: 5.8147 - coeff_determination: 0.6707
Epoch 60/100
24/24 [==============================] - 14s 573ms/step - loss: 50.8520 - mae: 5.6517 - coeff_determination: 0.6844
Epoch 61/100
24/24 [==============================] - 13s 562ms/step - loss: 54.6315 - mae: 5.8883 - coeff_determination: 0.6630
Epoch 62/100
24/24 [==============================] - 14s 563ms/step - loss: 57.2338 - mae: 6.0196 - coeff_determination: 0.6464
Epoch 63/100
24/24 [==============================] - 14s 577ms/step - loss: 51.3617 - mae: 5.6944 - coeff_determination: 0.6847
Epoch 64/100
24/24 [==============================] - 14s 572ms/step - loss: 51.6126 - mae: 5.7220 - coeff_determination: 0.6809
Epoch 65/100
24/24 [==============================] - 13s 557ms/step - loss: 55.5970 - mae: 5.9508 - coeff_determination: 0.6554
Epoch 66/100
24/24 [==============================] - 13s 559ms/step - loss: 52.2611 - mae: 5.7694 - coeff_determination: 0.6757
Epoch 67/100
24/24 [==============================] - 13s 561ms/step - loss: 47.2018 - mae: 5.4381 - coeff_determination: 0.7075
Epoch 68/100
24/24 [==============================] - 13s 558ms/step - loss: 53.1889 - mae: 5.7641 - coeff_determination: 0.6720
Epoch 69/100
24/24 [==============================] - 14s 564ms/step - loss: 61.6170 - mae: 6.2547 - coeff_determination: 0.6218
Epoch 70/100
24/24 [==============================] - 13s 556ms/step - loss: 48.2338 - mae: 5.5102 - coeff_determination: 0.7014
Epoch 71/100
24/24 [==============================] - 13s 558ms/step - loss: 46.2073 - mae: 5.4454 - coeff_determination: 0.7124
Epoch 72/100
24/24 [==============================] - 13s 556ms/step - loss: 54.6752 - mae: 5.9138 - coeff_determination: 0.6552
Epoch 73/100
24/24 [==============================] - 14s 568ms/step - loss: 76.4114 - mae: 6.9817 - coeff_determination: 0.5265
Epoch 74/100
24/24 [==============================] - 14s 566ms/step - loss: 49.2928 - mae: 5.5434 - coeff_determination: 0.6972
Epoch 75/100
24/24 [==============================] - 13s 561ms/step - loss: 74.3165 - mae: 6.9844 - coeff_determination: 0.5398
Epoch 76/100
24/24 [==============================] - 14s 574ms/step - loss: 67.5699 - mae: 6.6087 - coeff_determination: 0.5839
Epoch 77/100
24/24 [==============================] - 13s 559ms/step - loss: 46.7596 - mae: 5.4532 - coeff_determination: 0.7079
Epoch 78/100
24/24 [==============================] - 14s 566ms/step - loss: 61.4404 - mae: 6.2910 - coeff_determination: 0.6199
Epoch 79/100
24/24 [==============================] - 14s 567ms/step - loss: 66.6403 - mae: 6.4684 - coeff_determination: 0.5935
Epoch 80/100
24/24 [==============================] - 14s 563ms/step - loss: 64.0176 - mae: 6.3767 - coeff_determination: 0.6059
Epoch 81/100
24/24 [==============================] - 14s 571ms/step - loss: 45.5800 - mae: 5.3345 - coeff_determination: 0.7192
Epoch 82/100
24/24 [==============================] - 14s 569ms/step - loss: 42.2491 - mae: 5.1673 - coeff_determination: 0.7379
Epoch 83/100
24/24 [==============================] - 14s 566ms/step - loss: 43.0121 - mae: 5.2194 - coeff_determination: 0.7363
Epoch 84/100
24/24 [==============================] - 13s 559ms/step - loss: 40.6901 - mae: 5.0535 - coeff_determination: 0.7497
Epoch 85/100
24/24 [==============================] - 14s 566ms/step - loss: 49.3637 - mae: 5.5591 - coeff_determination: 0.6952
Epoch 86/100
24/24 [==============================] - 14s 564ms/step - loss: 43.4568 - mae: 5.2247 - coeff_determination: 0.7321
Epoch 87/100
24/24 [==============================] - 14s 575ms/step - loss: 44.6758 - mae: 5.2645 - coeff_determination: 0.7262
Epoch 88/100
24/24 [==============================] - 14s 569ms/step - loss: 45.4618 - mae: 5.3270 - coeff_determination: 0.7215
Epoch 89/100
24/24 [==============================] - 13s 558ms/step - loss: 42.7576 - mae: 5.1559 - coeff_determination: 0.7369
Epoch 90/100
24/24 [==============================] - 14s 565ms/step - loss: 40.8059 - mae: 5.0443 - coeff_determination: 0.7480
Epoch 91/100
24/24 [==============================] - 14s 572ms/step - loss: 46.3766 - mae: 5.4431 - coeff_determination: 0.7115
Epoch 92/100
24/24 [==============================] - 14s 568ms/step - loss: 41.4182 - mae: 5.0834 - coeff_determination: 0.7432
Epoch 93/100
24/24 [==============================] - 13s 560ms/step - loss: 48.2857 - mae: 5.5344 - coeff_determination: 0.7014
Epoch 94/100
24/24 [==============================] - 13s 558ms/step - loss: 62.7329 - mae: 6.3113 - coeff_determination: 0.6131
Epoch 95/100
24/24 [==============================] - 13s 558ms/step - loss: 56.6203 - mae: 5.9459 - coeff_determination: 0.6560
Epoch 96/100
24/24 [==============================] - 13s 558ms/step - loss: 41.6108 - mae: 5.1382 - coeff_determination: 0.7421
Epoch 97/100
24/24 [==============================] - 14s 580ms/step - loss: 43.8006 - mae: 5.2312 - coeff_determination: 0.7307
Epoch 98/100
24/24 [==============================] - 13s 560ms/step - loss: 48.9174 - mae: 5.5606 - coeff_determination: 0.6984
Epoch 99/100
24/24 [==============================] - 13s 562ms/step - loss: 47.1173 - mae: 5.4283 - coeff_determination: 0.7127
Epoch 100/100
24/24 [==============================] - 14s 568ms/step - loss: 42.6673 - mae: 5.1619 - coeff_determination: 0.7384
---------<TEST RESULTS FOR CLIENT client_1 ; USING LOCAL MODEL>-----------
R^2 0.406362364589821
Mean squared error 96.40476118993271
Mean absolute error 7.66770565521449
Huber loss 12.724429100888916
SCALING FACTOR : 0.5
--------<TEST RESULTS AFTER ROUND 0 ; USING GLOBAL MODEL>---------
R^2 -1.714
Mean squared error 440.672
Mean absolute error 18.943
Huber loss 20.111
==============================================================
==============================================================
---------<STARTING TRAINING FOR ROUND 1>-----------
---------<STARTING TRAINING FOR CLIENT client_2>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 16s 656ms/step - loss: 537.4771 - mae: 17.9297 - coeff_determination: -2.3635
Epoch 2/100
24/24 [==============================] - 14s 583ms/step - loss: 136.2800 - mae: 9.4214 - coeff_determination: 0.1640
Epoch 3/100
24/24 [==============================] - 14s 584ms/step - loss: 104.9561 - mae: 8.2436 - coeff_determination: 0.3518
Epoch 4/100
24/24 [==============================] - 14s 574ms/step - loss: 90.8756 - mae: 7.5710 - coeff_determination: 0.4442
Epoch 5/100
24/24 [==============================] - 14s 568ms/step - loss: 87.7680 - mae: 7.5262 - coeff_determination: 0.4645
Epoch 6/100
24/24 [==============================] - 14s 572ms/step - loss: 75.1600 - mae: 6.9360 - coeff_determination: 0.5380
Epoch 7/100
24/24 [==============================] - 14s 566ms/step - loss: 71.1198 - mae: 6.7723 - coeff_determination: 0.5639
Epoch 8/100
24/24 [==============================] - 13s 561ms/step - loss: 67.4269 - mae: 6.5766 - coeff_determination: 0.5825
Epoch 9/100
24/24 [==============================] - 14s 572ms/step - loss: 65.3083 - mae: 6.4562 - coeff_determination: 0.5930
Epoch 10/100
24/24 [==============================] - 14s 569ms/step - loss: 60.7811 - mae: 6.2109 - coeff_determination: 0.6260
Epoch 11/100
24/24 [==============================] - 13s 562ms/step - loss: 61.8680 - mae: 6.2703 - coeff_determination: 0.6207
Epoch 12/100
24/24 [==============================] - 13s 561ms/step - loss: 58.2057 - mae: 6.0421 - coeff_determination: 0.6436
Epoch 13/100
24/24 [==============================] - 13s 560ms/step - loss: 61.1608 - mae: 6.2428 - coeff_determination: 0.6235
Epoch 14/100
24/24 [==============================] - 14s 580ms/step - loss: 62.4358 - mae: 6.3293 - coeff_determination: 0.6158
Epoch 15/100
24/24 [==============================] - 13s 560ms/step - loss: 58.0028 - mae: 6.0838 - coeff_determination: 0.6422
Epoch 16/100
24/24 [==============================] - 14s 568ms/step - loss: 57.9846 - mae: 6.0854 - coeff_determination: 0.6395
Epoch 17/100
24/24 [==============================] - 13s 561ms/step - loss: 57.4747 - mae: 6.0142 - coeff_determination: 0.6510
Epoch 18/100
24/24 [==============================] - 14s 563ms/step - loss: 66.9797 - mae: 6.5195 - coeff_determination: 0.5896
Epoch 19/100
24/24 [==============================] - 14s 573ms/step - loss: 54.9632 - mae: 5.8923 - coeff_determination: 0.6654
Epoch 20/100
24/24 [==============================] - 14s 563ms/step - loss: 55.7292 - mae: 5.9047 - coeff_determination: 0.6572
Epoch 21/100
24/24 [==============================] - 14s 565ms/step - loss: 55.3862 - mae: 5.8930 - coeff_determination: 0.6615
Epoch 22/100
24/24 [==============================] - 14s 573ms/step - loss: 54.2379 - mae: 5.8049 - coeff_determination: 0.6671
Epoch 23/100
24/24 [==============================] - 14s 563ms/step - loss: 50.2729 - mae: 5.6270 - coeff_determination: 0.6908
Epoch 24/100
24/24 [==============================] - 14s 571ms/step - loss: 49.3548 - mae: 5.6025 - coeff_determination: 0.6952
Epoch 25/100
24/24 [==============================] - 14s 566ms/step - loss: 50.2880 - mae: 5.6133 - coeff_determination: 0.6902
Epoch 26/100
24/24 [==============================] - 14s 576ms/step - loss: 73.1999 - mae: 6.8347 - coeff_determination: 0.5496
Epoch 27/100
24/24 [==============================] - 14s 563ms/step - loss: 56.6209 - mae: 5.9368 - coeff_determination: 0.6550
Epoch 28/100
24/24 [==============================] - 14s 589ms/step - loss: 50.4019 - mae: 5.6147 - coeff_determination: 0.6927
Epoch 29/100
24/24 [==============================] - 14s 568ms/step - loss: 51.3893 - mae: 5.6870 - coeff_determination: 0.6851
Epoch 30/100
24/24 [==============================] - 13s 561ms/step - loss: 53.1564 - mae: 5.8913 - coeff_determination: 0.6645
Epoch 31/100
24/24 [==============================] - 14s 566ms/step - loss: 65.1074 - mae: 6.3650 - coeff_determination: 0.6057
Epoch 32/100
24/24 [==============================] - 14s 571ms/step - loss: 47.8692 - mae: 5.4751 - coeff_determination: 0.7066
Epoch 33/100
24/24 [==============================] - 14s 569ms/step - loss: 56.9134 - mae: 6.0410 - coeff_determination: 0.6518
Epoch 34/100
24/24 [==============================] - 14s 565ms/step - loss: 45.3325 - mae: 5.3477 - coeff_determination: 0.7205
Epoch 35/100
24/24 [==============================] - 14s 564ms/step - loss: 55.5724 - mae: 5.9444 - coeff_determination: 0.6588
Epoch 36/100
24/24 [==============================] - 14s 563ms/step - loss: 49.8616 - mae: 5.6304 - coeff_determination: 0.6961
Epoch 37/100
24/24 [==============================] - 14s 572ms/step - loss: 62.4263 - mae: 6.3316 - coeff_determination: 0.6170
Epoch 38/100
24/24 [==============================] - 14s 567ms/step - loss: 47.3218 - mae: 5.4359 - coeff_determination: 0.7116
Epoch 39/100
24/24 [==============================] - 14s 563ms/step - loss: 42.6007 - mae: 5.1817 - coeff_determination: 0.7370
Epoch 40/100
24/24 [==============================] - 14s 564ms/step - loss: 49.6426 - mae: 5.5135 - coeff_determination: 0.6977
Epoch 41/100
24/24 [==============================] - 14s 568ms/step - loss: 43.9769 - mae: 5.2838 - coeff_determination: 0.7321
Epoch 42/100
24/24 [==============================] - 14s 578ms/step - loss: 39.2311 - mae: 4.9695 - coeff_determination: 0.7567
Epoch 43/100
24/24 [==============================] - 14s 569ms/step - loss: 42.4253 - mae: 5.1299 - coeff_determination: 0.7409
Epoch 44/100
24/24 [==============================] - 14s 564ms/step - loss: 44.4676 - mae: 5.2866 - coeff_determination: 0.7235
Epoch 45/100
24/24 [==============================] - 14s 566ms/step - loss: 49.0922 - mae: 5.4876 - coeff_determination: 0.7016
Epoch 46/100
24/24 [==============================] - 14s 578ms/step - loss: 46.5204 - mae: 5.4404 - coeff_determination: 0.7091
Epoch 47/100
24/24 [==============================] - 14s 570ms/step - loss: 52.3367 - mae: 5.8119 - coeff_determination: 0.6756
Epoch 48/100
24/24 [==============================] - 14s 565ms/step - loss: 55.8746 - mae: 5.9817 - coeff_determination: 0.6537
Epoch 49/100
24/24 [==============================] - 14s 573ms/step - loss: 40.1482 - mae: 5.0415 - coeff_determination: 0.7523
Epoch 50/100
24/24 [==============================] - 14s 565ms/step - loss: 40.8935 - mae: 5.0307 - coeff_determination: 0.7505
Epoch 51/100
24/24 [==============================] - 14s 572ms/step - loss: 35.7164 - mae: 4.7083 - coeff_determination: 0.7802
Epoch 52/100
24/24 [==============================] - 14s 574ms/step - loss: 36.9060 - mae: 4.7898 - coeff_determination: 0.7744
Epoch 53/100
24/24 [==============================] - 13s 562ms/step - loss: 41.1888 - mae: 5.0998 - coeff_determination: 0.7474
Epoch 54/100
24/24 [==============================] - 14s 564ms/step - loss: 54.6324 - mae: 5.8378 - coeff_determination: 0.6674
Epoch 55/100
24/24 [==============================] - 14s 565ms/step - loss: 59.6346 - mae: 6.1893 - coeff_determination: 0.6344
Epoch 56/100
24/24 [==============================] - 14s 569ms/step - loss: 51.3169 - mae: 5.7196 - coeff_determination: 0.6871
Epoch 57/100
24/24 [==============================] - 14s 573ms/step - loss: 43.6510 - mae: 5.2162 - coeff_determination: 0.7318
Epoch 58/100
24/24 [==============================] - 13s 560ms/step - loss: 39.1219 - mae: 4.9174 - coeff_determination: 0.7597
Epoch 59/100
24/24 [==============================] - 14s 569ms/step - loss: 36.0779 - mae: 4.7111 - coeff_determination: 0.7795
Epoch 60/100
24/24 [==============================] - 14s 573ms/step - loss: 36.3587 - mae: 4.7845 - coeff_determination: 0.7745
Epoch 61/100
24/24 [==============================] - 14s 569ms/step - loss: 42.3346 - mae: 5.1691 - coeff_determination: 0.7401
Epoch 62/100
24/24 [==============================] - 14s 564ms/step - loss: 53.7700 - mae: 5.8370 - coeff_determination: 0.6603
Epoch 63/100
24/24 [==============================] - 14s 564ms/step - loss: 44.1751 - mae: 5.2740 - coeff_determination: 0.7293
Epoch 64/100
24/24 [==============================] - 14s 564ms/step - loss: 40.4511 - mae: 5.0435 - coeff_determination: 0.7540
Epoch 65/100
24/24 [==============================] - 14s 570ms/step - loss: 42.2555 - mae: 5.2265 - coeff_determination: 0.7301
Epoch 66/100
24/24 [==============================] - 14s 564ms/step - loss: 42.2233 - mae: 5.1071 - coeff_determination: 0.7418
Epoch 67/100
24/24 [==============================] - 14s 563ms/step - loss: 52.3242 - mae: 5.7934 - coeff_determination: 0.6784
Epoch 68/100
24/24 [==============================] - 14s 580ms/step - loss: 48.5072 - mae: 5.5029 - coeff_determination: 0.7045
Epoch 69/100
24/24 [==============================] - 14s 570ms/step - loss: 38.4143 - mae: 4.8911 - coeff_determination: 0.7642
Epoch 70/100
24/24 [==============================] - 14s 574ms/step - loss: 55.9277 - mae: 5.8966 - coeff_determination: 0.6574
Epoch 71/100
24/24 [==============================] - 14s 578ms/step - loss: 45.8047 - mae: 5.3688 - coeff_determination: 0.7185
Epoch 72/100
24/24 [==============================] - 14s 566ms/step - loss: 39.9531 - mae: 5.0245 - coeff_determination: 0.7535
Epoch 73/100
24/24 [==============================] - 14s 568ms/step - loss: 42.9223 - mae: 5.2805 - coeff_determination: 0.7287
Epoch 74/100
24/24 [==============================] - 14s 567ms/step - loss: 63.9085 - mae: 6.4293 - coeff_determination: 0.6086
Epoch 75/100
24/24 [==============================] - 14s 569ms/step - loss: 56.6572 - mae: 5.9593 - coeff_determination: 0.6548
Epoch 76/100
24/24 [==============================] - 13s 562ms/step - loss: 37.0824 - mae: 4.8175 - coeff_determination: 0.7709
Epoch 77/100
24/24 [==============================] - 14s 565ms/step - loss: 41.8014 - mae: 5.2241 - coeff_determination: 0.7323
Epoch 78/100
24/24 [==============================] - 14s 566ms/step - loss: 52.3047 - mae: 5.7544 - coeff_determination: 0.6815
Epoch 79/100
24/24 [==============================] - 14s 574ms/step - loss: 37.2337 - mae: 4.8187 - coeff_determination: 0.7713
Epoch 80/100
24/24 [==============================] - 14s 563ms/step - loss: 32.4823 - mae: 4.4788 - coeff_determination: 0.8007
Epoch 81/100
24/24 [==============================] - 14s 564ms/step - loss: 33.6569 - mae: 4.5692 - coeff_determination: 0.7926
Epoch 82/100
24/24 [==============================] - 14s 565ms/step - loss: 31.1727 - mae: 4.4392 - coeff_determination: 0.8054
Epoch 83/100
24/24 [==============================] - 13s 561ms/step - loss: 34.8724 - mae: 4.7187 - coeff_determination: 0.7813
Epoch 84/100
24/24 [==============================] - 14s 569ms/step - loss: 39.2713 - mae: 4.9842 - coeff_determination: 0.7591
Epoch 85/100
24/24 [==============================] - 14s 565ms/step - loss: 39.8905 - mae: 5.0425 - coeff_determination: 0.7511
Epoch 86/100
24/24 [==============================] - 14s 563ms/step - loss: 55.0361 - mae: 6.0333 - coeff_determination: 0.6575
Epoch 87/100
24/24 [==============================] - 14s 566ms/step - loss: 45.1727 - mae: 5.3718 - coeff_determination: 0.7208
Epoch 88/100
24/24 [==============================] - 14s 571ms/step - loss: 35.6896 - mae: 4.7487 - coeff_determination: 0.7766
Epoch 89/100
24/24 [==============================] - 14s 565ms/step - loss: 34.1873 - mae: 4.6322 - coeff_determination: 0.7872
Epoch 90/100
24/24 [==============================] - 14s 570ms/step - loss: 31.9717 - mae: 4.4698 - coeff_determination: 0.8032
Epoch 91/100
24/24 [==============================] - 14s 569ms/step - loss: 34.1630 - mae: 4.6074 - coeff_determination: 0.7901
Epoch 92/100
24/24 [==============================] - 14s 566ms/step - loss: 33.9865 - mae: 4.6173 - coeff_determination: 0.7896
Epoch 93/100
24/24 [==============================] - 14s 579ms/step - loss: 38.8151 - mae: 4.9584 - coeff_determination: 0.7620
Epoch 94/100
24/24 [==============================] - 14s 596ms/step - loss: 37.0256 - mae: 4.8404 - coeff_determination: 0.7721
Epoch 95/100
24/24 [==============================] - 14s 570ms/step - loss: 36.0704 - mae: 4.7818 - coeff_determination: 0.7790
Epoch 96/100
24/24 [==============================] - 14s 568ms/step - loss: 41.8156 - mae: 5.1079 - coeff_determination: 0.7418
Epoch 97/100
24/24 [==============================] - 14s 563ms/step - loss: 50.2939 - mae: 5.6273 - coeff_determination: 0.6964
Epoch 98/100
24/24 [==============================] - 14s 570ms/step - loss: 39.2196 - mae: 5.0426 - coeff_determination: 0.7583
Epoch 99/100
24/24 [==============================] - 14s 566ms/step - loss: 34.3815 - mae: 4.6108 - coeff_determination: 0.7917
Epoch 100/100
24/24 [==============================] - 14s 566ms/step - loss: 33.7560 - mae: 4.5570 - coeff_determination: 0.7943
---------<TEST RESULTS FOR CLIENT client_2 ; USING LOCAL MODEL>-----------
R^2 0.482545568216178
Mean squared error 84.03286440611731
Mean absolute error 7.025532622724218
Huber loss 12.555097339318015
SCALING FACTOR : 0.5
---------<STARTING TRAINING FOR CLIENT client_1>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 15s 618ms/step - loss: 545.9411 - mae: 18.4728 - coeff_determination: -2.4000
Epoch 2/100
24/24 [==============================] - 14s 572ms/step - loss: 145.6075 - mae: 9.7342 - coeff_determination: 0.1079
Epoch 3/100
24/24 [==============================] - 13s 562ms/step - loss: 108.0071 - mae: 8.4323 - coeff_determination: 0.3341
Epoch 4/100
24/24 [==============================] - 14s 564ms/step - loss: 87.6495 - mae: 7.5284 - coeff_determination: 0.4651
Epoch 5/100
24/24 [==============================] - 14s 564ms/step - loss: 77.0831 - mae: 7.0558 - coeff_determination: 0.5280
Epoch 6/100
24/24 [==============================] - 14s 564ms/step - loss: 67.7602 - mae: 6.6091 - coeff_determination: 0.5825
Epoch 7/100
24/24 [==============================] - 14s 567ms/step - loss: 62.9761 - mae: 6.3311 - coeff_determination: 0.6130
Epoch 8/100
24/24 [==============================] - 14s 563ms/step - loss: 60.5960 - mae: 6.1984 - coeff_determination: 0.6272
Epoch 9/100
24/24 [==============================] - 14s 563ms/step - loss: 62.9866 - mae: 6.3451 - coeff_determination: 0.6130
Epoch 10/100
24/24 [==============================] - 14s 565ms/step - loss: 61.7320 - mae: 6.2082 - coeff_determination: 0.6218
Epoch 11/100
24/24 [==============================] - 14s 570ms/step - loss: 57.8789 - mae: 6.0904 - coeff_determination: 0.6414
Epoch 12/100
24/24 [==============================] - 14s 575ms/step - loss: 76.1949 - mae: 6.9457 - coeff_determination: 0.5354
Epoch 13/100
24/24 [==============================] - 14s 572ms/step - loss: 58.8784 - mae: 6.1337 - coeff_determination: 0.6323
Epoch 14/100
24/24 [==============================] - 13s 561ms/step - loss: 56.5429 - mae: 5.9941 - coeff_determination: 0.6522
Epoch 15/100
24/24 [==============================] - 14s 567ms/step - loss: 56.8999 - mae: 5.9856 - coeff_determination: 0.6530
Epoch 16/100
24/24 [==============================] - 14s 587ms/step - loss: 46.2366 - mae: 5.4037 - coeff_determination: 0.7159
Epoch 17/100
24/24 [==============================] - 14s 568ms/step - loss: 54.8861 - mae: 5.8763 - coeff_determination: 0.6617
Epoch 18/100
24/24 [==============================] - 13s 562ms/step - loss: 62.9320 - mae: 6.4555 - coeff_determination: 0.6011
Epoch 19/100
24/24 [==============================] - 14s 571ms/step - loss: 60.6822 - mae: 6.1755 - coeff_determination: 0.6291
Epoch 20/100
24/24 [==============================] - 14s 567ms/step - loss: 48.9590 - mae: 5.5854 - coeff_determination: 0.6982
Epoch 21/100
24/24 [==============================] - 14s 567ms/step - loss: 51.2277 - mae: 5.6986 - coeff_determination: 0.6813
Epoch 22/100
24/24 [==============================] - 14s 567ms/step - loss: 63.4048 - mae: 6.4105 - coeff_determination: 0.6118
Epoch 23/100
24/24 [==============================] - 13s 561ms/step - loss: 53.0428 - mae: 5.7385 - coeff_determination: 0.6769
Epoch 24/100
24/24 [==============================] - 14s 566ms/step - loss: 45.8776 - mae: 5.4074 - coeff_determination: 0.7133
Epoch 25/100
24/24 [==============================] - 14s 571ms/step - loss: 48.4214 - mae: 5.5447 - coeff_determination: 0.7014
Epoch 26/100
24/24 [==============================] - 13s 562ms/step - loss: 52.5228 - mae: 5.7577 - coeff_determination: 0.6754
Epoch 27/100
24/24 [==============================] - 14s 568ms/step - loss: 43.7590 - mae: 5.2551 - coeff_determination: 0.7266
Epoch 28/100
24/24 [==============================] - 13s 560ms/step - loss: 45.9565 - mae: 5.3743 - coeff_determination: 0.7120
Epoch 29/100
24/24 [==============================] - 14s 568ms/step - loss: 46.2375 - mae: 5.3572 - coeff_determination: 0.7147
Epoch 30/100
24/24 [==============================] - 14s 584ms/step - loss: 43.0914 - mae: 5.1928 - coeff_determination: 0.7326
Epoch 31/100
24/24 [==============================] - 14s 591ms/step - loss: 50.4859 - mae: 5.6878 - coeff_determination: 0.6852
Epoch 32/100
24/24 [==============================] - 15s 606ms/step - loss: 65.9775 - mae: 6.5842 - coeff_determination: 0.5850
Epoch 33/100
24/24 [==============================] - 18s 733ms/step - loss: 46.0747 - mae: 5.3473 - coeff_determination: 0.7179
Epoch 34/100
24/24 [==============================] - 16s 687ms/step - loss: 43.6432 - mae: 5.2580 - coeff_determination: 0.7303
Epoch 35/100
24/24 [==============================] - 17s 695ms/step - loss: 53.3566 - mae: 5.8291 - coeff_determination: 0.6716
Epoch 36/100
24/24 [==============================] - 16s 653ms/step - loss: 45.4892 - mae: 5.3596 - coeff_determination: 0.7193
Epoch 37/100
24/24 [==============================] - 15s 644ms/step - loss: 40.1841 - mae: 5.0197 - coeff_determination: 0.7521
Epoch 38/100
24/24 [==============================] - 15s 622ms/step - loss: 38.4629 - mae: 4.9182 - coeff_determination: 0.7621
Epoch 39/100
24/24 [==============================] - 14s 594ms/step - loss: 41.2194 - mae: 5.0803 - coeff_determination: 0.7451
Epoch 40/100
24/24 [==============================] - 19s 791ms/step - loss: 52.1095 - mae: 5.7356 - coeff_determination: 0.6836
Epoch 41/100
24/24 [==============================] - 19s 798ms/step - loss: 41.2715 - mae: 5.1171 - coeff_determination: 0.7464
Epoch 42/100
24/24 [==============================] - 17s 697ms/step - loss: 52.0261 - mae: 5.7564 - coeff_determination: 0.6820
Epoch 43/100
24/24 [==============================] - 17s 708ms/step - loss: 47.3917 - mae: 5.4577 - coeff_determination: 0.7102
Epoch 44/100
24/24 [==============================] - 17s 722ms/step - loss: 40.6865 - mae: 5.0774 - coeff_determination: 0.7482
Epoch 45/100
24/24 [==============================] - 16s 662ms/step - loss: 38.9236 - mae: 4.9436 - coeff_determination: 0.7583
Epoch 46/100
24/24 [==============================] - 14s 588ms/step - loss: 40.0867 - mae: 4.9893 - coeff_determination: 0.7552
Epoch 47/100
24/24 [==============================] - 14s 564ms/step - loss: 56.8062 - mae: 6.0145 - coeff_determination: 0.6532
Epoch 48/100
24/24 [==============================] - 14s 595ms/step - loss: 47.8062 - mae: 5.4270 - coeff_determination: 0.7066
Epoch 49/100
24/24 [==============================] - 16s 671ms/step - loss: 40.4778 - mae: 5.0677 - coeff_determination: 0.7465
Epoch 50/100
24/24 [==============================] - 17s 719ms/step - loss: 43.8355 - mae: 5.2317 - coeff_determination: 0.7297
Epoch 51/100
24/24 [==============================] - 14s 585ms/step - loss: 39.2834 - mae: 4.9627 - coeff_determination: 0.7579
Epoch 52/100
24/24 [==============================] - 14s 595ms/step - loss: 34.3782 - mae: 4.6384 - coeff_determination: 0.7868
Epoch 53/100
24/24 [==============================] - 14s 595ms/step - loss: 34.9788 - mae: 4.6813 - coeff_determination: 0.7819
Epoch 54/100
24/24 [==============================] - 17s 695ms/step - loss: 43.9222 - mae: 5.2921 - coeff_determination: 0.7280
Epoch 55/100
24/24 [==============================] - 15s 622ms/step - loss: 61.0561 - mae: 6.2363 - coeff_determination: 0.6267
Epoch 56/100
24/24 [==============================] - 19s 783ms/step - loss: 42.6011 - mae: 5.1899 - coeff_determination: 0.7391
Epoch 57/100
24/24 [==============================] - 18s 746ms/step - loss: 40.2439 - mae: 5.0699 - coeff_determination: 0.7483
Epoch 58/100
24/24 [==============================] - 14s 603ms/step - loss: 37.7077 - mae: 4.8809 - coeff_determination: 0.7659
Epoch 59/100
24/24 [==============================] - 16s 683ms/step - loss: 36.6920 - mae: 4.8190 - coeff_determination: 0.7724
Epoch 60/100
24/24 [==============================] - 14s 574ms/step - loss: 37.4427 - mae: 4.8621 - coeff_determination: 0.7704
Epoch 61/100
24/24 [==============================] - 15s 610ms/step - loss: 35.8232 - mae: 4.7498 - coeff_determination: 0.7754
Epoch 62/100
24/24 [==============================] - 14s 586ms/step - loss: 35.9207 - mae: 4.7554 - coeff_determination: 0.7787
Epoch 63/100
24/24 [==============================] - 14s 582ms/step - loss: 37.4438 - mae: 4.8297 - coeff_determination: 0.7683
Epoch 64/100
24/24 [==============================] - 13s 546ms/step - loss: 36.3515 - mae: 4.8090 - coeff_determination: 0.7707
Epoch 65/100
24/24 [==============================] - 13s 544ms/step - loss: 39.2431 - mae: 4.9387 - coeff_determination: 0.7587
Epoch 66/100
24/24 [==============================] - 13s 540ms/step - loss: 43.5379 - mae: 5.2521 - coeff_determination: 0.7309
Epoch 67/100
24/24 [==============================] - 13s 545ms/step - loss: 42.9499 - mae: 5.2162 - coeff_determination: 0.7370
Epoch 68/100
24/24 [==============================] - 13s 546ms/step - loss: 33.7855 - mae: 4.6249 - coeff_determination: 0.7923
Epoch 69/100
24/24 [==============================] - 16s 659ms/step - loss: 35.6318 - mae: 4.7271 - coeff_determination: 0.7794
Epoch 70/100
24/24 [==============================] - 15s 614ms/step - loss: 50.1619 - mae: 5.6150 - coeff_determination: 0.6937
Epoch 71/100
24/24 [==============================] - 14s 600ms/step - loss: 38.2332 - mae: 4.9455 - coeff_determination: 0.7633
Epoch 72/100
24/24 [==============================] - 14s 580ms/step - loss: 32.8553 - mae: 4.5091 - coeff_determination: 0.7985
Epoch 73/100
24/24 [==============================] - 15s 621ms/step - loss: 36.3326 - mae: 4.7647 - coeff_determination: 0.7769
Epoch 74/100
24/24 [==============================] - 15s 622ms/step - loss: 46.1152 - mae: 5.3731 - coeff_determination: 0.7186
Epoch 75/100
24/24 [==============================] - 15s 616ms/step - loss: 37.3895 - mae: 4.8529 - coeff_determination: 0.7671
Epoch 76/100
24/24 [==============================] - 14s 600ms/step - loss: 33.8630 - mae: 4.5653 - coeff_determination: 0.7921
Epoch 77/100
24/24 [==============================] - 15s 613ms/step - loss: 40.5823 - mae: 5.0621 - coeff_determination: 0.7484
Epoch 78/100
24/24 [==============================] - 14s 584ms/step - loss: 39.6204 - mae: 4.9652 - coeff_determination: 0.7578
Epoch 79/100
24/24 [==============================] - 15s 605ms/step - loss: 33.9230 - mae: 4.6145 - coeff_determination: 0.7895
Epoch 80/100
24/24 [==============================] - 14s 604ms/step - loss: 36.7407 - mae: 4.7832 - coeff_determination: 0.7745
Epoch 81/100
24/24 [==============================] - 13s 540ms/step - loss: 39.1562 - mae: 5.0212 - coeff_determination: 0.7569
Epoch 82/100
24/24 [==============================] - 13s 554ms/step - loss: 33.9642 - mae: 4.6426 - coeff_determination: 0.7880
Epoch 83/100
24/24 [==============================] - 14s 577ms/step - loss: 34.7475 - mae: 4.6973 - coeff_determination: 0.7814
Epoch 84/100
24/24 [==============================] - 13s 541ms/step - loss: 39.9146 - mae: 5.0274 - coeff_determination: 0.7524
Epoch 85/100
24/24 [==============================] - 14s 584ms/step - loss: 44.1041 - mae: 5.3263 - coeff_determination: 0.7258
Epoch 86/100
24/24 [==============================] - 16s 660ms/step - loss: 50.3346 - mae: 5.6976 - coeff_determination: 0.6800
Epoch 87/100
24/24 [==============================] - 16s 654ms/step - loss: 42.7625 - mae: 5.1689 - coeff_determination: 0.7369
Epoch 88/100
24/24 [==============================] - 16s 651ms/step - loss: 33.6605 - mae: 4.6098 - coeff_determination: 0.7897
Epoch 89/100
24/24 [==============================] - 14s 600ms/step - loss: 45.1377 - mae: 5.4625 - coeff_determination: 0.7107
Epoch 90/100
24/24 [==============================] - 14s 589ms/step - loss: 68.5184 - mae: 6.7087 - coeff_determination: 0.5771
Epoch 91/100
24/24 [==============================] - 14s 570ms/step - loss: 37.4705 - mae: 4.8327 - coeff_determination: 0.7707
Epoch 92/100
24/24 [==============================] - 14s 573ms/step - loss: 42.7464 - mae: 5.2068 - coeff_determination: 0.7390
Epoch 93/100
24/24 [==============================] - 14s 573ms/step - loss: 41.9122 - mae: 5.1457 - coeff_determination: 0.7428
Epoch 94/100
24/24 [==============================] - 14s 583ms/step - loss: 32.8664 - mae: 4.5285 - coeff_determination: 0.7952
Epoch 95/100
24/24 [==============================] - 14s 577ms/step - loss: 45.2379 - mae: 5.3470 - coeff_determination: 0.7256
Epoch 96/100
24/24 [==============================] - 14s 565ms/step - loss: 32.5724 - mae: 4.5392 - coeff_determination: 0.7993
Epoch 97/100
24/24 [==============================] - 14s 580ms/step - loss: 38.2933 - mae: 4.9151 - coeff_determination: 0.7647
Epoch 98/100
24/24 [==============================] - 14s 571ms/step - loss: 31.8788 - mae: 4.4872 - coeff_determination: 0.8031
Epoch 99/100
24/24 [==============================] - 14s 574ms/step - loss: 37.3455 - mae: 4.8572 - coeff_determination: 0.7670
Epoch 100/100
24/24 [==============================] - 14s 571ms/step - loss: 33.9425 - mae: 4.5805 - coeff_determination: 0.7922
---------<TEST RESULTS FOR CLIENT client_1 ; USING LOCAL MODEL>-----------
R^2 0.3444732880346729
Mean squared error 106.45533967362668
Mean absolute error 8.01487638927819
Huber loss 13.245920631346657
SCALING FACTOR : 0.5
--------<TEST RESULTS AFTER ROUND 1 ; USING GLOBAL MODEL>---------
R^2 0.288
Mean squared error 115.678
Mean absolute error 8.746
Huber loss 13.345
==============================================================
==============================================================
---------<STARTING TRAINING FOR ROUND 2>-----------
---------<STARTING TRAINING FOR CLIENT client_2>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 19s 798ms/step - loss: 1174.4077 - mae: 27.0728 - coeff_determination: -6.3737
Epoch 2/100
24/24 [==============================] - 15s 618ms/step - loss: 148.8867 - mae: 9.8263 - coeff_determination: 0.0825
Epoch 3/100
24/24 [==============================] - 16s 655ms/step - loss: 119.2245 - mae: 8.7649 - coeff_determination: 0.2687
Epoch 4/100
24/24 [==============================] - 15s 639ms/step - loss: 108.1891 - mae: 8.3972 - coeff_determination: 0.3320
Epoch 5/100
24/24 [==============================] - 14s 568ms/step - loss: 98.6630 - mae: 7.9696 - coeff_determination: 0.3937
Epoch 6/100
24/24 [==============================] - 16s 686ms/step - loss: 86.3969 - mae: 7.4170 - coeff_determination: 0.4704
Epoch 7/100
24/24 [==============================] - 14s 602ms/step - loss: 79.2152 - mae: 7.1932 - coeff_determination: 0.5116
Epoch 8/100
24/24 [==============================] - 17s 698ms/step - loss: 69.7429 - mae: 6.6637 - coeff_determination: 0.5698
Epoch 9/100
24/24 [==============================] - 16s 649ms/step - loss: 69.0966 - mae: 6.6641 - coeff_determination: 0.5746
Epoch 10/100
24/24 [==============================] - 14s 572ms/step - loss: 66.2579 - mae: 6.5664 - coeff_determination: 0.5867
Epoch 11/100
24/24 [==============================] - 16s 676ms/step - loss: 61.0535 - mae: 6.2978 - coeff_determination: 0.6129
Epoch 12/100
24/24 [==============================] - 14s 584ms/step - loss: 60.7027 - mae: 6.2600 - coeff_determination: 0.6216
Epoch 13/100
24/24 [==============================] - 15s 605ms/step - loss: 58.8285 - mae: 6.1289 - coeff_determination: 0.6381
Epoch 14/100
24/24 [==============================] - 14s 595ms/step - loss: 71.5050 - mae: 6.7400 - coeff_determination: 0.5611
Epoch 15/100
24/24 [==============================] - 14s 563ms/step - loss: 56.5920 - mae: 5.9774 - coeff_determination: 0.6519
Epoch 16/100
24/24 [==============================] - 13s 550ms/step - loss: 52.1846 - mae: 5.7616 - coeff_determination: 0.6758
Epoch 17/100
24/24 [==============================] - 14s 571ms/step - loss: 50.3746 - mae: 5.6165 - coeff_determination: 0.6917
Epoch 18/100
24/24 [==============================] - 14s 603ms/step - loss: 78.5968 - mae: 7.0518 - coeff_determination: 0.5199
Epoch 19/100
24/24 [==============================] - 14s 583ms/step - loss: 60.6540 - mae: 6.1444 - coeff_determination: 0.6329
Epoch 20/100
24/24 [==============================] - 14s 603ms/step - loss: 54.4609 - mae: 5.8617 - coeff_determination: 0.6664
Epoch 21/100
24/24 [==============================] - 16s 648ms/step - loss: 55.7908 - mae: 5.9842 - coeff_determination: 0.6540
Epoch 22/100
24/24 [==============================] - 14s 565ms/step - loss: 50.1740 - mae: 5.6224 - coeff_determination: 0.6901
Epoch 23/100
24/24 [==============================] - 14s 594ms/step - loss: 46.7167 - mae: 5.4213 - coeff_determination: 0.7131
Epoch 24/100
24/24 [==============================] - 14s 567ms/step - loss: 44.5747 - mae: 5.2884 - coeff_determination: 0.7254
Epoch 25/100
24/24 [==============================] - 13s 551ms/step - loss: 52.4337 - mae: 5.7151 - coeff_determination: 0.6771
Epoch 26/100
24/24 [==============================] - 13s 543ms/step - loss: 48.6880 - mae: 5.5465 - coeff_determination: 0.7007
Epoch 27/100
24/24 [==============================] - 13s 549ms/step - loss: 50.3802 - mae: 5.7141 - coeff_determination: 0.6790
Epoch 28/100
24/24 [==============================] - 13s 545ms/step - loss: 48.3335 - mae: 5.5020 - coeff_determination: 0.7038
Epoch 29/100
24/24 [==============================] - 13s 541ms/step - loss: 43.2643 - mae: 5.2046 - coeff_determination: 0.7343
Epoch 30/100
24/24 [==============================] - 13s 547ms/step - loss: 40.8621 - mae: 5.0708 - coeff_determination: 0.7488
Epoch 31/100
24/24 [==============================] - 13s 541ms/step - loss: 44.9081 - mae: 5.3248 - coeff_determination: 0.7234
Epoch 32/100
24/24 [==============================] - 13s 543ms/step - loss: 47.1914 - mae: 5.4522 - coeff_determination: 0.7128
Epoch 33/100
24/24 [==============================] - 13s 545ms/step - loss: 40.5209 - mae: 5.0653 - coeff_determination: 0.7471
Epoch 34/100
24/24 [==============================] - 13s 543ms/step - loss: 43.6116 - mae: 5.2285 - coeff_determination: 0.7305
Epoch 35/100
24/24 [==============================] - 14s 567ms/step - loss: 45.2657 - mae: 5.4618 - coeff_determination: 0.7073
Epoch 36/100
24/24 [==============================] - 13s 549ms/step - loss: 79.4435 - mae: 7.1014 - coeff_determination: 0.5193
Epoch 37/100
24/24 [==============================] - 13s 555ms/step - loss: 50.5312 - mae: 5.6466 - coeff_determination: 0.6888
Epoch 38/100
24/24 [==============================] - 14s 601ms/step - loss: 41.8824 - mae: 5.1242 - coeff_determination: 0.7431
Epoch 39/100
24/24 [==============================] - 13s 547ms/step - loss: 41.0010 - mae: 5.0446 - coeff_determination: 0.7497
Epoch 40/100
24/24 [==============================] - 13s 557ms/step - loss: 42.0472 - mae: 5.1872 - coeff_determination: 0.7379
Epoch 41/100
24/24 [==============================] - 13s 539ms/step - loss: 44.5904 - mae: 5.3014 - coeff_determination: 0.7254
Epoch 42/100
24/24 [==============================] - 13s 545ms/step - loss: 38.1725 - mae: 4.9218 - coeff_determination: 0.7643
Epoch 43/100
24/24 [==============================] - 13s 554ms/step - loss: 38.6398 - mae: 4.9012 - coeff_determination: 0.7627
Epoch 44/100
24/24 [==============================] - 13s 561ms/step - loss: 52.4668 - mae: 5.7417 - coeff_determination: 0.6758
Epoch 45/100
24/24 [==============================] - 13s 546ms/step - loss: 42.3481 - mae: 5.1668 - coeff_determination: 0.7393
Epoch 46/100
24/24 [==============================] - 13s 548ms/step - loss: 42.1631 - mae: 5.1849 - coeff_determination: 0.7396
Epoch 47/100
24/24 [==============================] - 13s 541ms/step - loss: 37.5658 - mae: 4.8617 - coeff_determination: 0.7685
Epoch 48/100
24/24 [==============================] - 13s 545ms/step - loss: 36.7717 - mae: 4.7939 - coeff_determination: 0.7730
Epoch 49/100
24/24 [==============================] - 13s 545ms/step - loss: 35.2387 - mae: 4.7174 - coeff_determination: 0.7810
Epoch 50/100
24/24 [==============================] - 13s 544ms/step - loss: 36.2900 - mae: 4.7955 - coeff_determination: 0.7751
Epoch 51/100
24/24 [==============================] - 13s 544ms/step - loss: 35.2685 - mae: 4.6960 - coeff_determination: 0.7831
Epoch 52/100
24/24 [==============================] - 13s 550ms/step - loss: 59.7703 - mae: 6.1332 - coeff_determination: 0.6308
Epoch 53/100
24/24 [==============================] - 13s 547ms/step - loss: 54.8909 - mae: 5.8632 - coeff_determination: 0.6654
Epoch 54/100
24/24 [==============================] - 13s 547ms/step - loss: 39.0674 - mae: 4.9258 - coeff_determination: 0.7618
Epoch 55/100
24/24 [==============================] - 13s 558ms/step - loss: 41.3913 - mae: 5.1988 - coeff_determination: 0.7326
Epoch 56/100
24/24 [==============================] - 13s 542ms/step - loss: 42.2893 - mae: 5.1692 - coeff_determination: 0.7383
Epoch 57/100
24/24 [==============================] - 13s 547ms/step - loss: 35.9434 - mae: 4.7181 - coeff_determination: 0.7798
Epoch 58/100
24/24 [==============================] - 13s 552ms/step - loss: 32.4031 - mae: 4.4907 - coeff_determination: 0.7991
Epoch 59/100
24/24 [==============================] - 13s 554ms/step - loss: 40.7075 - mae: 5.1921 - coeff_determination: 0.7328
Epoch 60/100
24/24 [==============================] - 13s 547ms/step - loss: 54.8588 - mae: 5.9242 - coeff_determination: 0.6602
Epoch 61/100
24/24 [==============================] - 13s 545ms/step - loss: 50.6277 - mae: 5.6247 - coeff_determination: 0.6900
Epoch 62/100
24/24 [==============================] - 13s 560ms/step - loss: 36.4156 - mae: 4.8120 - coeff_determination: 0.7734
Epoch 63/100
24/24 [==============================] - 13s 547ms/step - loss: 41.0081 - mae: 5.0555 - coeff_determination: 0.7482
Epoch 64/100
24/24 [==============================] - 13s 550ms/step - loss: 35.0756 - mae: 4.6926 - coeff_determination: 0.7859
Epoch 65/100
24/24 [==============================] - 13s 541ms/step - loss: 68.6810 - mae: 6.5738 - coeff_determination: 0.5888
Epoch 66/100
24/24 [==============================] - 13s 552ms/step - loss: 47.3895 - mae: 5.5233 - coeff_determination: 0.7051
Epoch 67/100
24/24 [==============================] - 13s 543ms/step - loss: 39.2661 - mae: 4.9560 - coeff_determination: 0.7579
Epoch 68/100
24/24 [==============================] - 13s 553ms/step - loss: 49.1859 - mae: 5.6210 - coeff_determination: 0.6961
Epoch 69/100
24/24 [==============================] - 13s 547ms/step - loss: 46.0070 - mae: 5.3809 - coeff_determination: 0.7184
Epoch 70/100
24/24 [==============================] - 13s 545ms/step - loss: 33.8239 - mae: 4.5802 - coeff_determination: 0.7917
Epoch 71/100
24/24 [==============================] - 13s 549ms/step - loss: 39.1629 - mae: 4.9541 - coeff_determination: 0.7602
Epoch 72/100
24/24 [==============================] - 13s 544ms/step - loss: 35.6393 - mae: 4.7290 - coeff_determination: 0.7819
Epoch 73/100
24/24 [==============================] - 13s 547ms/step - loss: 34.8496 - mae: 4.6976 - coeff_determination: 0.7840
Epoch 74/100
24/24 [==============================] - 13s 550ms/step - loss: 35.2743 - mae: 4.6924 - coeff_determination: 0.7838
Epoch 75/100
24/24 [==============================] - 13s 546ms/step - loss: 38.7579 - mae: 4.9083 - coeff_determination: 0.7642
Epoch 76/100
24/24 [==============================] - 13s 545ms/step - loss: 29.4470 - mae: 4.2627 - coeff_determination: 0.8174
Epoch 77/100
24/24 [==============================] - 13s 548ms/step - loss: 30.8307 - mae: 4.3793 - coeff_determination: 0.8098
Epoch 78/100
24/24 [==============================] - 13s 546ms/step - loss: 42.4887 - mae: 5.1624 - coeff_determination: 0.7362
Epoch 79/100
24/24 [==============================] - 13s 546ms/step - loss: 37.8681 - mae: 4.8590 - coeff_determination: 0.7703
Epoch 80/100
24/24 [==============================] - 13s 551ms/step - loss: 36.1088 - mae: 4.7351 - coeff_determination: 0.7786
Epoch 81/100
24/24 [==============================] - 13s 546ms/step - loss: 43.3007 - mae: 5.2097 - coeff_determination: 0.7322
Epoch 82/100
24/24 [==============================] - 13s 548ms/step - loss: 38.3661 - mae: 4.8978 - coeff_determination: 0.7648
Epoch 83/100
24/24 [==============================] - 15s 613ms/step - loss: 35.9656 - mae: 4.7638 - coeff_determination: 0.7782
Epoch 84/100
24/24 [==============================] - 14s 596ms/step - loss: 36.0389 - mae: 4.7723 - coeff_determination: 0.7745
Epoch 85/100
24/24 [==============================] - 14s 603ms/step - loss: 40.1839 - mae: 5.0341 - coeff_determination: 0.7554
Epoch 86/100
24/24 [==============================] - 14s 580ms/step - loss: 45.5318 - mae: 5.4461 - coeff_determination: 0.7182
Epoch 87/100
24/24 [==============================] - 14s 578ms/step - loss: 34.6397 - mae: 4.6606 - coeff_determination: 0.7892
Epoch 88/100
24/24 [==============================] - 14s 576ms/step - loss: 31.9764 - mae: 4.4969 - coeff_determination: 0.8029
Epoch 89/100
24/24 [==============================] - 14s 574ms/step - loss: 33.3526 - mae: 4.5576 - coeff_determination: 0.7937
Epoch 90/100
24/24 [==============================] - 14s 573ms/step - loss: 29.9725 - mae: 4.3085 - coeff_determination: 0.8150
Epoch 91/100
24/24 [==============================] - 14s 598ms/step - loss: 31.2265 - mae: 4.3795 - coeff_determination: 0.8079
Epoch 92/100
24/24 [==============================] - 14s 570ms/step - loss: 31.1116 - mae: 4.4121 - coeff_determination: 0.8095
Epoch 93/100
24/24 [==============================] - 14s 574ms/step - loss: 33.2310 - mae: 4.5329 - coeff_determination: 0.7965
Epoch 94/100
24/24 [==============================] - 14s 585ms/step - loss: 28.5876 - mae: 4.2096 - coeff_determination: 0.8245
Epoch 95/100
24/24 [==============================] - 14s 572ms/step - loss: 31.4475 - mae: 4.4596 - coeff_determination: 0.8047
Epoch 96/100
24/24 [==============================] - 14s 584ms/step - loss: 40.2228 - mae: 5.0898 - coeff_determination: 0.7455
Epoch 97/100
24/24 [==============================] - 14s 577ms/step - loss: 65.1875 - mae: 6.4942 - coeff_determination: 0.5983
Epoch 98/100
24/24 [==============================] - 14s 575ms/step - loss: 34.2815 - mae: 4.6734 - coeff_determination: 0.7856
Epoch 99/100
24/24 [==============================] - 14s 575ms/step - loss: 33.5146 - mae: 4.5596 - coeff_determination: 0.7945
Epoch 100/100
24/24 [==============================] - 14s 584ms/step - loss: 36.9996 - mae: 4.8964 - coeff_determination: 0.7627
---------<TEST RESULTS FOR CLIENT client_2 ; USING LOCAL MODEL>-----------
R^2 0.4855842596234714
Mean squared error 83.53939111201325
Mean absolute error 7.166759803575582
Huber loss 12.811988372425846
SCALING FACTOR : 0.5
---------<STARTING TRAINING FOR CLIENT client_1>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 16s 670ms/step - loss: 1141.5711 - mae: 26.4209 - coeff_determination: -6.1255
Epoch 2/100
24/24 [==============================] - 14s 579ms/step - loss: 152.8567 - mae: 10.0136 - coeff_determination: 0.0568
Epoch 3/100
24/24 [==============================] - 14s 575ms/step - loss: 124.1517 - mae: 9.0166 - coeff_determination: 0.2316
Epoch 4/100
24/24 [==============================] - 14s 584ms/step - loss: 104.1374 - mae: 8.1772 - coeff_determination: 0.3615
Epoch 5/100
24/24 [==============================] - 14s 577ms/step - loss: 99.4276 - mae: 8.0553 - coeff_determination: 0.3866
Epoch 6/100
24/24 [==============================] - 14s 578ms/step - loss: 87.6220 - mae: 7.4871 - coeff_determination: 0.4632
Epoch 7/100
24/24 [==============================] - 14s 574ms/step - loss: 82.6900 - mae: 7.2835 - coeff_determination: 0.4888
Epoch 8/100
24/24 [==============================] - 14s 577ms/step - loss: 70.7269 - mae: 6.6759 - coeff_determination: 0.5656
Epoch 9/100
24/24 [==============================] - 14s 581ms/step - loss: 79.1898 - mae: 7.0817 - coeff_determination: 0.5147
Epoch 10/100
24/24 [==============================] - 14s 576ms/step - loss: 68.8601 - mae: 6.6645 - coeff_determination: 0.5710
Epoch 11/100
24/24 [==============================] - 14s 573ms/step - loss: 72.3992 - mae: 6.8385 - coeff_determination: 0.5534
Epoch 12/100
24/24 [==============================] - 14s 580ms/step - loss: 75.2127 - mae: 6.9623 - coeff_determination: 0.5414
Epoch 13/100
24/24 [==============================] - 14s 576ms/step - loss: 56.5393 - mae: 6.0092 - coeff_determination: 0.6509
Epoch 14/100
24/24 [==============================] - 13s 528ms/step - loss: 52.6820 - mae: 5.7920 - coeff_determination: 0.6746
Epoch 15/100
24/24 [==============================] - 13s 534ms/step - loss: 53.4076 - mae: 5.8180 - coeff_determination: 0.6691
Epoch 16/100
24/24 [==============================] - 14s 579ms/step - loss: 50.5588 - mae: 5.6710 - coeff_determination: 0.6898
Epoch 17/100
24/24 [==============================] - 14s 574ms/step - loss: 55.4182 - mae: 5.9867 - coeff_determination: 0.6516
Epoch 18/100
24/24 [==============================] - 14s 576ms/step - loss: 52.5030 - mae: 5.7410 - coeff_determination: 0.6782
Epoch 19/100
24/24 [==============================] - 14s 575ms/step - loss: 55.6697 - mae: 5.9356 - coeff_determination: 0.6566
Epoch 20/100
24/24 [==============================] - 14s 576ms/step - loss: 57.5836 - mae: 6.0625 - coeff_determination: 0.6445
Epoch 21/100
24/24 [==============================] - 14s 572ms/step - loss: 46.6851 - mae: 5.4169 - coeff_determination: 0.7108
Epoch 22/100
24/24 [==============================] - 14s 577ms/step - loss: 60.0677 - mae: 6.1720 - coeff_determination: 0.6369
Epoch 23/100
24/24 [==============================] - 14s 600ms/step - loss: 44.9119 - mae: 5.3483 - coeff_determination: 0.7221
Epoch 24/100
24/24 [==============================] - 14s 575ms/step - loss: 49.8125 - mae: 5.6069 - coeff_determination: 0.6955
Epoch 25/100
24/24 [==============================] - 14s 588ms/step - loss: 52.1567 - mae: 5.7469 - coeff_determination: 0.6781
Epoch 26/100
24/24 [==============================] - 14s 582ms/step - loss: 42.7896 - mae: 5.1730 - coeff_determination: 0.7352
Epoch 27/100
24/24 [==============================] - 14s 582ms/step - loss: 42.5707 - mae: 5.1862 - coeff_determination: 0.7371
Epoch 28/100
24/24 [==============================] - 14s 584ms/step - loss: 51.6433 - mae: 5.7339 - coeff_determination: 0.6788
Epoch 29/100
24/24 [==============================] - 14s 589ms/step - loss: 47.4561 - mae: 5.4626 - coeff_determination: 0.7065
Epoch 30/100
24/24 [==============================] - 14s 574ms/step - loss: 40.3901 - mae: 5.0555 - coeff_determination: 0.7509
Epoch 31/100
24/24 [==============================] - 14s 577ms/step - loss: 56.4542 - mae: 6.0345 - coeff_determination: 0.6518
Epoch 32/100
24/24 [==============================] - 14s 577ms/step - loss: 48.7929 - mae: 5.5343 - coeff_determination: 0.6972
Epoch 33/100
24/24 [==============================] - 14s 578ms/step - loss: 42.0081 - mae: 5.1625 - coeff_determination: 0.7419
Epoch 34/100
24/24 [==============================] - 14s 587ms/step - loss: 40.0510 - mae: 5.0500 - coeff_determination: 0.7492
Epoch 35/100
24/24 [==============================] - 14s 587ms/step - loss: 38.7021 - mae: 4.9186 - coeff_determination: 0.7600
Epoch 36/100
24/24 [==============================] - 14s 577ms/step - loss: 51.2331 - mae: 5.6978 - coeff_determination: 0.6842
Epoch 37/100
24/24 [==============================] - 14s 577ms/step - loss: 41.8790 - mae: 5.1669 - coeff_determination: 0.7407
Epoch 38/100
24/24 [==============================] - 14s 579ms/step - loss: 42.4045 - mae: 5.1965 - coeff_determination: 0.7363
Epoch 39/100
24/24 [==============================] - 14s 573ms/step - loss: 36.5002 - mae: 4.8098 - coeff_determination: 0.7725
Epoch 40/100
24/24 [==============================] - 14s 578ms/step - loss: 37.4473 - mae: 4.8746 - coeff_determination: 0.7666
Epoch 41/100
24/24 [==============================] - 14s 577ms/step - loss: 39.0919 - mae: 4.9473 - coeff_determination: 0.7598
Epoch 42/100
24/24 [==============================] - 14s 582ms/step - loss: 36.9057 - mae: 4.8198 - coeff_determination: 0.7703
Epoch 43/100
24/24 [==============================] - 14s 583ms/step - loss: 42.4705 - mae: 5.1666 - coeff_determination: 0.7371
Epoch 44/100
24/24 [==============================] - 14s 575ms/step - loss: 54.2301 - mae: 5.8843 - coeff_determination: 0.6635
Epoch 45/100
24/24 [==============================] - 14s 580ms/step - loss: 47.9779 - mae: 5.5399 - coeff_determination: 0.6977
Epoch 46/100
24/24 [==============================] - 14s 578ms/step - loss: 40.4867 - mae: 5.0962 - coeff_determination: 0.7440
Epoch 47/100
24/24 [==============================] - 14s 579ms/step - loss: 43.7501 - mae: 5.2146 - coeff_determination: 0.7326
Epoch 48/100
24/24 [==============================] - 14s 579ms/step - loss: 35.9673 - mae: 4.7420 - coeff_determination: 0.7788
Epoch 49/100
24/24 [==============================] - 18s 739ms/step - loss: 46.2015 - mae: 5.4458 - coeff_determination: 0.7085
Epoch 50/100
24/24 [==============================] - 14s 603ms/step - loss: 49.5634 - mae: 5.6395 - coeff_determination: 0.6918
Epoch 51/100
24/24 [==============================] - 15s 638ms/step - loss: 51.3972 - mae: 5.6585 - coeff_determination: 0.6823
Epoch 52/100
24/24 [==============================] - 15s 633ms/step - loss: 40.3427 - mae: 5.0128 - coeff_determination: 0.7532
Epoch 53/100
24/24 [==============================] - 13s 544ms/step - loss: 39.1933 - mae: 4.9519 - coeff_determination: 0.7612
Epoch 54/100
24/24 [==============================] - 17s 702ms/step - loss: 35.0141 - mae: 4.7013 - coeff_determination: 0.7802
Epoch 55/100
24/24 [==============================] - 14s 598ms/step - loss: 33.0745 - mae: 4.5662 - coeff_determination: 0.7953
Epoch 56/100
24/24 [==============================] - 17s 688ms/step - loss: 36.8873 - mae: 4.8620 - coeff_determination: 0.7719
Epoch 57/100
24/24 [==============================] - 17s 705ms/step - loss: 39.8476 - mae: 5.0527 - coeff_determination: 0.7492
Epoch 58/100
24/24 [==============================] - 17s 714ms/step - loss: 54.8125 - mae: 5.9266 - coeff_determination: 0.6637
Epoch 59/100
24/24 [==============================] - 15s 630ms/step - loss: 35.2315 - mae: 4.6927 - coeff_determination: 0.7799
Epoch 60/100
24/24 [==============================] - 14s 581ms/step - loss: 36.7938 - mae: 4.8262 - coeff_determination: 0.7729
Epoch 61/100
24/24 [==============================] - 14s 598ms/step - loss: 40.9202 - mae: 5.1409 - coeff_determination: 0.7424
Epoch 62/100
24/24 [==============================] - 14s 590ms/step - loss: 47.6010 - mae: 5.5060 - coeff_determination: 0.7054
Epoch 63/100
24/24 [==============================] - 17s 726ms/step - loss: 32.0861 - mae: 4.4874 - coeff_determination: 0.8017
Epoch 64/100
24/24 [==============================] - 16s 683ms/step - loss: 46.7470 - mae: 5.4235 - coeff_determination: 0.7170
Epoch 65/100
24/24 [==============================] - 13s 546ms/step - loss: 38.6596 - mae: 4.9225 - coeff_determination: 0.7631
Epoch 66/100
24/24 [==============================] - 15s 607ms/step - loss: 39.3497 - mae: 4.9802 - coeff_determination: 0.7555
Epoch 67/100
24/24 [==============================] - 18s 760ms/step - loss: 35.9601 - mae: 4.7624 - coeff_determination: 0.7744
Epoch 68/100
24/24 [==============================] - 14s 588ms/step - loss: 36.8186 - mae: 4.7642 - coeff_determination: 0.7752
Epoch 69/100
24/24 [==============================] - 15s 622ms/step - loss: 32.4865 - mae: 4.5050 - coeff_determination: 0.7999
Epoch 70/100
24/24 [==============================] - 13s 556ms/step - loss: 30.5051 - mae: 4.3644 - coeff_determination: 0.8110
Epoch 71/100
24/24 [==============================] - 14s 582ms/step - loss: 34.1785 - mae: 4.6512 - coeff_determination: 0.7883
Epoch 72/100
24/24 [==============================] - 24s 983ms/step - loss: 43.6993 - mae: 5.2011 - coeff_determination: 0.7331
Epoch 73/100
24/24 [==============================] - 23s 966ms/step - loss: 48.9992 - mae: 5.5405 - coeff_determination: 0.7031
Epoch 74/100
24/24 [==============================] - 17s 705ms/step - loss: 36.2409 - mae: 4.7846 - coeff_determination: 0.7777
Epoch 75/100
24/24 [==============================] - 16s 657ms/step - loss: 37.2511 - mae: 4.8438 - coeff_determination: 0.7707
Epoch 76/100
24/24 [==============================] - 17s 715ms/step - loss: 32.6323 - mae: 4.5339 - coeff_determination: 0.7954
Epoch 77/100
24/24 [==============================] - 16s 650ms/step - loss: 35.6714 - mae: 4.7688 - coeff_determination: 0.7756
Epoch 78/100
24/24 [==============================] - 15s 620ms/step - loss: 36.7774 - mae: 4.8165 - coeff_determination: 0.7732
Epoch 79/100
24/24 [==============================] - 16s 667ms/step - loss: 33.2150 - mae: 4.5490 - coeff_determination: 0.7955
Epoch 80/100
24/24 [==============================] - 16s 665ms/step - loss: 30.6302 - mae: 4.4031 - coeff_determination: 0.8117
Epoch 81/100
24/24 [==============================] - 17s 699ms/step - loss: 30.2946 - mae: 4.3357 - coeff_determination: 0.8123
Epoch 82/100
24/24 [==============================] - 14s 602ms/step - loss: 31.4617 - mae: 4.5006 - coeff_determination: 0.8010
Epoch 83/100
24/24 [==============================] - 13s 558ms/step - loss: 37.2033 - mae: 4.9019 - coeff_determination: 0.7650
Epoch 84/100
24/24 [==============================] - 18s 743ms/step - loss: 31.2802 - mae: 4.4123 - coeff_determination: 0.8081
Epoch 85/100
24/24 [==============================] - 17s 697ms/step - loss: 34.3818 - mae: 4.5931 - coeff_determination: 0.7903
Epoch 86/100
24/24 [==============================] - 15s 627ms/step - loss: 44.2251 - mae: 5.2733 - coeff_determination: 0.7305
Epoch 87/100
24/24 [==============================] - 15s 639ms/step - loss: 35.1733 - mae: 4.6774 - coeff_determination: 0.7830
Epoch 88/100
24/24 [==============================] - 13s 553ms/step - loss: 44.5560 - mae: 5.2967 - coeff_determination: 0.7286
Epoch 89/100
24/24 [==============================] - 13s 551ms/step - loss: 28.2234 - mae: 4.1867 - coeff_determination: 0.8243
Epoch 90/100
24/24 [==============================] - 13s 543ms/step - loss: 31.0560 - mae: 4.4144 - coeff_determination: 0.8089
Epoch 91/100
24/24 [==============================] - 14s 567ms/step - loss: 30.0287 - mae: 4.3561 - coeff_determination: 0.8139
Epoch 92/100
24/24 [==============================] - 13s 553ms/step - loss: 32.3961 - mae: 4.5145 - coeff_determination: 0.7980
Epoch 93/100
24/24 [==============================] - 13s 553ms/step - loss: 29.7120 - mae: 4.2794 - coeff_determination: 0.8181
Epoch 94/100
24/24 [==============================] - 13s 549ms/step - loss: 31.7320 - mae: 4.4373 - coeff_determination: 0.8034
Epoch 95/100
24/24 [==============================] - 15s 613ms/step - loss: 35.1161 - mae: 4.7122 - coeff_determination: 0.7842
Epoch 96/100
24/24 [==============================] - 16s 648ms/step - loss: 40.6502 - mae: 5.0100 - coeff_determination: 0.7517
Epoch 97/100
24/24 [==============================] - 14s 578ms/step - loss: 30.9931 - mae: 4.3869 - coeff_determination: 0.8100
Epoch 98/100
24/24 [==============================] - 16s 677ms/step - loss: 37.4913 - mae: 4.8725 - coeff_determination: 0.7703
Epoch 99/100
24/24 [==============================] - 14s 576ms/step - loss: 30.3058 - mae: 4.3163 - coeff_determination: 0.8146
Epoch 100/100
24/24 [==============================] - 17s 710ms/step - loss: 29.1149 - mae: 4.2909 - coeff_determination: 0.8205
---------<TEST RESULTS FOR CLIENT client_1 ; USING LOCAL MODEL>-----------
R^2 0.4186883990860364
Mean squared error 94.40305452386956
Mean absolute error 7.493844823442635
Huber loss 12.762769329916468
SCALING FACTOR : 0.5
--------<TEST RESULTS AFTER ROUND 2 ; USING GLOBAL MODEL>---------
R^2 -0.306
Mean squared error 212.104
Mean absolute error 12.374
Huber loss 15.609
==============================================================
==============================================================
---------<STARTING TRAINING FOR ROUND 3>-----------
---------<STARTING TRAINING FOR CLIENT client_1>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 19s 774ms/step - loss: 632.6967 - mae: 19.4913 - coeff_determination: -2.8320
Epoch 2/100
24/24 [==============================] - 17s 722ms/step - loss: 111.4825 - mae: 8.4696 - coeff_determination: 0.3185
Epoch 3/100
24/24 [==============================] - 15s 640ms/step - loss: 81.9352 - mae: 7.2510 - coeff_determination: 0.4949
Epoch 4/100
24/24 [==============================] - 18s 747ms/step - loss: 70.8381 - mae: 6.7660 - coeff_determination: 0.5619
Epoch 5/100
24/24 [==============================] - 19s 798ms/step - loss: 66.1571 - mae: 6.5005 - coeff_determination: 0.5921
Epoch 6/100
24/24 [==============================] - 20s 819ms/step - loss: 60.2343 - mae: 6.2095 - coeff_determination: 0.6301
Epoch 7/100
24/24 [==============================] - 21s 896ms/step - loss: 51.6445 - mae: 5.7496 - coeff_determination: 0.6815
Epoch 8/100
24/24 [==============================] - 18s 765ms/step - loss: 53.6499 - mae: 5.8839 - coeff_determination: 0.6699
Epoch 9/100
24/24 [==============================] - 19s 805ms/step - loss: 52.1568 - mae: 5.7529 - coeff_determination: 0.6806
Epoch 10/100
24/24 [==============================] - 16s 670ms/step - loss: 46.5788 - mae: 5.4675 - coeff_determination: 0.7120
Epoch 11/100
24/24 [==============================] - 15s 618ms/step - loss: 51.7581 - mae: 5.7432 - coeff_determination: 0.6813
Epoch 12/100
24/24 [==============================] - 13s 544ms/step - loss: 40.6814 - mae: 5.0861 - coeff_determination: 0.7477
Epoch 13/100
24/24 [==============================] - 13s 538ms/step - loss: 41.9212 - mae: 5.1118 - coeff_determination: 0.7411
Epoch 14/100
24/24 [==============================] - 13s 552ms/step - loss: 40.7585 - mae: 5.0945 - coeff_determination: 0.7485
Epoch 15/100
24/24 [==============================] - 13s 541ms/step - loss: 48.3691 - mae: 5.4962 - coeff_determination: 0.7043
Epoch 16/100
24/24 [==============================] - 15s 630ms/step - loss: 43.4038 - mae: 5.2440 - coeff_determination: 0.7331
Epoch 17/100
24/24 [==============================] - 13s 546ms/step - loss: 38.9099 - mae: 4.9317 - coeff_determination: 0.7591
Epoch 18/100
24/24 [==============================] - 13s 541ms/step - loss: 53.7405 - mae: 5.8496 - coeff_determination: 0.6635
Epoch 19/100
24/24 [==============================] - 13s 552ms/step - loss: 49.9997 - mae: 5.6577 - coeff_determination: 0.6917
Epoch 20/100
24/24 [==============================] - 13s 541ms/step - loss: 40.9847 - mae: 5.0330 - coeff_determination: 0.7483
Epoch 21/100
24/24 [==============================] - 13s 554ms/step - loss: 35.5748 - mae: 4.7282 - coeff_determination: 0.7809
Epoch 22/100
24/24 [==============================] - 13s 541ms/step - loss: 34.8924 - mae: 4.6937 - coeff_determination: 0.7846
Epoch 23/100
24/24 [==============================] - 13s 539ms/step - loss: 36.0906 - mae: 4.7634 - coeff_determination: 0.7764
Epoch 24/100
24/24 [==============================] - 13s 538ms/step - loss: 40.3454 - mae: 5.0128 - coeff_determination: 0.7528
Epoch 25/100
24/24 [==============================] - 13s 543ms/step - loss: 38.0510 - mae: 4.9037 - coeff_determination: 0.7641
Epoch 26/100
24/24 [==============================] - 13s 560ms/step - loss: 37.2085 - mae: 4.8198 - coeff_determination: 0.7717
Epoch 27/100
24/24 [==============================] - 13s 537ms/step - loss: 48.6653 - mae: 5.5431 - coeff_determination: 0.6979
Epoch 28/100
24/24 [==============================] - 13s 547ms/step - loss: 42.5553 - mae: 5.1536 - coeff_determination: 0.7386
Epoch 29/100
24/24 [==============================] - 13s 540ms/step - loss: 50.5618 - mae: 5.6491 - coeff_determination: 0.6870
Epoch 30/100
24/24 [==============================] - 13s 541ms/step - loss: 48.1158 - mae: 5.5115 - coeff_determination: 0.7046
Epoch 31/100
24/24 [==============================] - 13s 547ms/step - loss: 34.8502 - mae: 4.6799 - coeff_determination: 0.7848
Epoch 32/100
24/24 [==============================] - 13s 538ms/step - loss: 35.5195 - mae: 4.7188 - coeff_determination: 0.7783
Epoch 33/100
24/24 [==============================] - 13s 542ms/step - loss: 32.7103 - mae: 4.5299 - coeff_determination: 0.7992
Epoch 34/100
24/24 [==============================] - 13s 544ms/step - loss: 34.4470 - mae: 4.6501 - coeff_determination: 0.7884
Epoch 35/100
24/24 [==============================] - 13s 546ms/step - loss: 33.0590 - mae: 4.5372 - coeff_determination: 0.7959
Epoch 36/100
24/24 [==============================] - 13s 537ms/step - loss: 33.0691 - mae: 4.5670 - coeff_determination: 0.7945
Epoch 37/100
24/24 [==============================] - 13s 541ms/step - loss: 37.0733 - mae: 4.8438 - coeff_determination: 0.7706
Epoch 38/100
24/24 [==============================] - 13s 545ms/step - loss: 46.4016 - mae: 5.4598 - coeff_determination: 0.7116
Epoch 39/100
24/24 [==============================] - 13s 542ms/step - loss: 36.6257 - mae: 4.7828 - coeff_determination: 0.7750
Epoch 40/100
24/24 [==============================] - 13s 539ms/step - loss: 35.5179 - mae: 4.7043 - coeff_determination: 0.7816
Epoch 41/100
24/24 [==============================] - 13s 538ms/step - loss: 36.4719 - mae: 4.7705 - coeff_determination: 0.7760
Epoch 42/100
24/24 [==============================] - 13s 544ms/step - loss: 33.3787 - mae: 4.5787 - coeff_determination: 0.7917
Epoch 43/100
24/24 [==============================] - 13s 542ms/step - loss: 33.6391 - mae: 4.5845 - coeff_determination: 0.7949
Epoch 44/100
24/24 [==============================] - 13s 557ms/step - loss: 33.5517 - mae: 4.5529 - coeff_determination: 0.7938
Epoch 45/100
24/24 [==============================] - 13s 541ms/step - loss: 35.6865 - mae: 4.7235 - coeff_determination: 0.7799
Epoch 46/100
24/24 [==============================] - 13s 539ms/step - loss: 34.7768 - mae: 4.6547 - coeff_determination: 0.7864
Epoch 47/100
24/24 [==============================] - 13s 546ms/step - loss: 43.4844 - mae: 5.2591 - coeff_determination: 0.7305
Epoch 48/100
24/24 [==============================] - 13s 547ms/step - loss: 36.3535 - mae: 4.7661 - coeff_determination: 0.7743
Epoch 49/100
24/24 [==============================] - 13s 545ms/step - loss: 47.7528 - mae: 5.5483 - coeff_determination: 0.6981
Epoch 50/100
24/24 [==============================] - 13s 541ms/step - loss: 40.6652 - mae: 5.0190 - coeff_determination: 0.7514
Epoch 51/100
24/24 [==============================] - 13s 540ms/step - loss: 31.3619 - mae: 4.4644 - coeff_determination: 0.8041
Epoch 52/100
24/24 [==============================] - 13s 546ms/step - loss: 37.2328 - mae: 4.8520 - coeff_determination: 0.7666
Epoch 53/100
24/24 [==============================] - 13s 540ms/step - loss: 35.6866 - mae: 4.7229 - coeff_determination: 0.7815
Epoch 54/100
24/24 [==============================] - 13s 538ms/step - loss: 30.3305 - mae: 4.3266 - coeff_determination: 0.8127
Epoch 55/100
24/24 [==============================] - 13s 544ms/step - loss: 45.9147 - mae: 5.4922 - coeff_determination: 0.7096
Epoch 56/100
24/24 [==============================] - 13s 542ms/step - loss: 71.8968 - mae: 6.8403 - coeff_determination: 0.5571
Epoch 57/100
24/24 [==============================] - 13s 549ms/step - loss: 43.8732 - mae: 5.2333 - coeff_determination: 0.7321
Epoch 58/100
24/24 [==============================] - 13s 539ms/step - loss: 31.7591 - mae: 4.4675 - coeff_determination: 0.8025
Epoch 59/100
24/24 [==============================] - 13s 539ms/step - loss: 37.0297 - mae: 4.8719 - coeff_determination: 0.7682
Epoch 60/100
24/24 [==============================] - 14s 576ms/step - loss: 30.3143 - mae: 4.3400 - coeff_determination: 0.8132
Epoch 61/100
24/24 [==============================] - 15s 616ms/step - loss: 31.4195 - mae: 4.4659 - coeff_determination: 0.8024
Epoch 62/100
24/24 [==============================] - 16s 651ms/step - loss: 38.3951 - mae: 4.9292 - coeff_determination: 0.7625
Epoch 63/100
24/24 [==============================] - 26s 1s/step - loss: 34.0109 - mae: 4.6167 - coeff_determination: 0.7888
Epoch 64/100
24/24 [==============================] - 19s 797ms/step - loss: 32.2508 - mae: 4.4581 - coeff_determination: 0.8033
Epoch 65/100
24/24 [==============================] - 13s 544ms/step - loss: 31.7706 - mae: 4.4971 - coeff_determination: 0.8006
Epoch 66/100
24/24 [==============================] - 13s 557ms/step - loss: 40.4330 - mae: 5.0511 - coeff_determination: 0.7509
Epoch 67/100
24/24 [==============================] - 16s 685ms/step - loss: 32.8441 - mae: 4.5402 - coeff_determination: 0.7986
Epoch 68/100
24/24 [==============================] - 15s 613ms/step - loss: 32.0159 - mae: 4.4357 - coeff_determination: 0.8044
Epoch 69/100
24/24 [==============================] - 15s 606ms/step - loss: 29.1881 - mae: 4.2990 - coeff_determination: 0.8169
Epoch 70/100
24/24 [==============================] - 14s 575ms/step - loss: 29.7508 - mae: 4.3384 - coeff_determination: 0.8157
Epoch 71/100
24/24 [==============================] - 15s 614ms/step - loss: 30.1321 - mae: 4.3422 - coeff_determination: 0.8132
Epoch 72/100
24/24 [==============================] - 14s 599ms/step - loss: 44.8803 - mae: 5.3611 - coeff_determination: 0.7244
Epoch 73/100
24/24 [==============================] - 15s 609ms/step - loss: 41.4342 - mae: 5.1898 - coeff_determination: 0.7384
Epoch 74/100
24/24 [==============================] - 15s 616ms/step - loss: 37.5555 - mae: 4.8258 - coeff_determination: 0.7700
Epoch 75/100
24/24 [==============================] - 14s 583ms/step - loss: 31.6825 - mae: 4.4398 - coeff_determination: 0.8048
Epoch 76/100
24/24 [==============================] - 14s 563ms/step - loss: 29.6231 - mae: 4.3235 - coeff_determination: 0.8150
Epoch 77/100
24/24 [==============================] - 13s 539ms/step - loss: 38.0054 - mae: 4.9101 - coeff_determination: 0.7615
Epoch 78/100
24/24 [==============================] - 13s 543ms/step - loss: 29.2914 - mae: 4.2570 - coeff_determination: 0.8173
Epoch 79/100
24/24 [==============================] - 13s 545ms/step - loss: 29.3603 - mae: 4.2687 - coeff_determination: 0.8202
Epoch 80/100
24/24 [==============================] - 13s 542ms/step - loss: 34.9191 - mae: 4.6960 - coeff_determination: 0.7829
Epoch 81/100
24/24 [==============================] - 14s 573ms/step - loss: 29.4405 - mae: 4.2465 - coeff_determination: 0.8208
Epoch 82/100
24/24 [==============================] - 14s 576ms/step - loss: 39.1546 - mae: 4.9690 - coeff_determination: 0.7608
Epoch 83/100
24/24 [==============================] - 13s 539ms/step - loss: 29.2735 - mae: 4.2912 - coeff_determination: 0.8177
Epoch 84/100
24/24 [==============================] - 14s 572ms/step - loss: 30.7059 - mae: 4.3999 - coeff_determination: 0.8081
Epoch 85/100
24/24 [==============================] - 13s 541ms/step - loss: 36.9898 - mae: 4.8712 - coeff_determination: 0.7690
Epoch 86/100
24/24 [==============================] - 14s 599ms/step - loss: 27.2188 - mae: 4.1305 - coeff_determination: 0.8321
Epoch 87/100
24/24 [==============================] - 16s 649ms/step - loss: 35.9500 - mae: 4.6899 - coeff_determination: 0.7804
Epoch 88/100
24/24 [==============================] - 14s 598ms/step - loss: 31.0481 - mae: 4.4137 - coeff_determination: 0.8084
Epoch 89/100
24/24 [==============================] - 14s 591ms/step - loss: 29.5415 - mae: 4.2772 - coeff_determination: 0.8184
Epoch 90/100
24/24 [==============================] - 14s 585ms/step - loss: 26.3064 - mae: 4.0285 - coeff_determination: 0.8369
Epoch 91/100
24/24 [==============================] - 14s 598ms/step - loss: 35.2868 - mae: 4.7243 - coeff_determination: 0.7804
Epoch 92/100
24/24 [==============================] - 16s 653ms/step - loss: 29.2936 - mae: 4.2795 - coeff_determination: 0.8188
Epoch 93/100
24/24 [==============================] - 13s 556ms/step - loss: 43.6324 - mae: 5.2297 - coeff_determination: 0.7318
Epoch 94/100
24/24 [==============================] - 13s 538ms/step - loss: 31.7996 - mae: 4.4558 - coeff_determination: 0.8033
Epoch 95/100
24/24 [==============================] - 13s 538ms/step - loss: 41.0148 - mae: 5.0630 - coeff_determination: 0.7496
Epoch 96/100
24/24 [==============================] - 13s 546ms/step - loss: 28.7464 - mae: 4.2359 - coeff_determination: 0.8221
Epoch 97/100
24/24 [==============================] - 14s 581ms/step - loss: 24.6877 - mae: 3.9248 - coeff_determination: 0.8469
Epoch 98/100
24/24 [==============================] - 13s 545ms/step - loss: 25.6757 - mae: 4.0040 - coeff_determination: 0.8428
Epoch 99/100
24/24 [==============================] - 14s 566ms/step - loss: 38.0524 - mae: 4.9388 - coeff_determination: 0.7617
Epoch 100/100
24/24 [==============================] - 13s 545ms/step - loss: 30.2674 - mae: 4.3359 - coeff_determination: 0.8145
---------<TEST RESULTS FOR CLIENT client_1 ; USING LOCAL MODEL>-----------
R^2 0.5032949233687738
Mean squared error 80.66323871358699
Mean absolute error 6.814059883657128
Huber loss 12.810701904219174
SCALING FACTOR : 0.5
---------<STARTING TRAINING FOR CLIENT client_2>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 19s 789ms/step - loss: 668.0346 - mae: 20.0541 - coeff_determination: -3.0394
Epoch 2/100
24/24 [==============================] - 15s 636ms/step - loss: 108.6632 - mae: 8.3106 - coeff_determination: 0.3407
Epoch 3/100
24/24 [==============================] - 16s 664ms/step - loss: 89.8321 - mae: 7.6551 - coeff_determination: 0.4498
Epoch 4/100
24/24 [==============================] - 14s 574ms/step - loss: 71.2844 - mae: 6.7745 - coeff_determination: 0.5647
Epoch 5/100
24/24 [==============================] - 14s 565ms/step - loss: 60.8778 - mae: 6.2313 - coeff_determination: 0.6285
Epoch 6/100
24/24 [==============================] - 16s 658ms/step - loss: 58.5745 - mae: 6.0911 - coeff_determination: 0.6446
Epoch 7/100
24/24 [==============================] - 14s 570ms/step - loss: 58.5439 - mae: 6.1109 - coeff_determination: 0.6392
Epoch 8/100
24/24 [==============================] - 14s 569ms/step - loss: 58.9473 - mae: 6.1730 - coeff_determination: 0.6381
Epoch 9/100
24/24 [==============================] - 14s 563ms/step - loss: 59.7279 - mae: 6.1763 - coeff_determination: 0.6347
Epoch 10/100
24/24 [==============================] - 13s 540ms/step - loss: 52.8676 - mae: 5.8472 - coeff_determination: 0.6704
Epoch 11/100
24/24 [==============================] - 13s 542ms/step - loss: 50.3546 - mae: 5.6600 - coeff_determination: 0.6924
Epoch 12/100
24/24 [==============================] - 13s 545ms/step - loss: 42.8542 - mae: 5.1776 - coeff_determination: 0.7382
Epoch 13/100
24/24 [==============================] - 13s 548ms/step - loss: 42.2026 - mae: 5.1532 - coeff_determination: 0.7415
Epoch 14/100
24/24 [==============================] - 13s 541ms/step - loss: 42.8295 - mae: 5.1942 - coeff_determination: 0.7390
Epoch 15/100
24/24 [==============================] - 13s 555ms/step - loss: 45.5356 - mae: 5.3343 - coeff_determination: 0.7218
Epoch 16/100
24/24 [==============================] - 13s 545ms/step - loss: 44.6095 - mae: 5.2895 - coeff_determination: 0.7274
Epoch 17/100
24/24 [==============================] - 13s 543ms/step - loss: 38.8354 - mae: 4.9454 - coeff_determination: 0.7622
Epoch 18/100
24/24 [==============================] - 13s 544ms/step - loss: 42.1253 - mae: 5.1379 - coeff_determination: 0.7413
Epoch 19/100
24/24 [==============================] - 13s 542ms/step - loss: 41.7088 - mae: 5.0955 - coeff_determination: 0.7443
Epoch 20/100
24/24 [==============================] - 13s 542ms/step - loss: 35.8496 - mae: 4.7476 - coeff_determination: 0.7791
Epoch 21/100
24/24 [==============================] - 13s 548ms/step - loss: 39.2296 - mae: 5.0065 - coeff_determination: 0.7558
Epoch 22/100
24/24 [==============================] - 13s 542ms/step - loss: 40.7932 - mae: 5.0630 - coeff_determination: 0.7478
Epoch 23/100
24/24 [==============================] - 13s 556ms/step - loss: 36.8419 - mae: 4.8206 - coeff_determination: 0.7728
Epoch 24/100
24/24 [==============================] - 15s 637ms/step - loss: 48.4674 - mae: 5.5872 - coeff_determination: 0.6937
Epoch 25/100
24/24 [==============================] - 14s 604ms/step - loss: 43.6223 - mae: 5.1944 - coeff_determination: 0.7335
Epoch 26/100
24/24 [==============================] - 15s 638ms/step - loss: 47.2014 - mae: 5.4426 - coeff_determination: 0.7095
Epoch 27/100
24/24 [==============================] - 16s 665ms/step - loss: 39.2335 - mae: 4.9552 - coeff_determination: 0.7578
Epoch 28/100
24/24 [==============================] - 15s 612ms/step - loss: 34.1115 - mae: 4.6079 - coeff_determination: 0.7902
Epoch 29/100
24/24 [==============================] - 16s 658ms/step - loss: 36.8874 - mae: 4.8213 - coeff_determination: 0.7726
Epoch 30/100
24/24 [==============================] - 14s 599ms/step - loss: 35.6842 - mae: 4.7323 - coeff_determination: 0.7800
Epoch 31/100
24/24 [==============================] - 15s 632ms/step - loss: 35.7944 - mae: 4.7177 - coeff_determination: 0.7804
Epoch 32/100
24/24 [==============================] - 13s 554ms/step - loss: 34.8597 - mae: 4.6373 - coeff_determination: 0.7854
Epoch 33/100
24/24 [==============================] - 14s 597ms/step - loss: 39.5588 - mae: 4.9823 - coeff_determination: 0.7587
Epoch 34/100
24/24 [==============================] - 15s 618ms/step - loss: 34.3567 - mae: 4.6668 - coeff_determination: 0.7885
Epoch 35/100
24/24 [==============================] - 14s 599ms/step - loss: 38.4747 - mae: 4.9355 - coeff_determination: 0.7624
Epoch 36/100
24/24 [==============================] - 14s 585ms/step - loss: 36.1648 - mae: 4.7463 - coeff_determination: 0.7752
Epoch 37/100
24/24 [==============================] - 14s 587ms/step - loss: 35.8854 - mae: 4.7627 - coeff_determination: 0.7799
Epoch 38/100
24/24 [==============================] - 14s 592ms/step - loss: 32.5338 - mae: 4.4860 - coeff_determination: 0.8008
Epoch 39/100
24/24 [==============================] - 14s 584ms/step - loss: 60.3886 - mae: 6.1868 - coeff_determination: 0.6292
Epoch 40/100
24/24 [==============================] - 14s 584ms/step - loss: 40.8877 - mae: 5.0726 - coeff_determination: 0.7490
Epoch 41/100
24/24 [==============================] - 14s 600ms/step - loss: 36.0714 - mae: 4.8096 - coeff_determination: 0.7727
Epoch 42/100
24/24 [==============================] - 14s 587ms/step - loss: 32.6446 - mae: 4.5304 - coeff_determination: 0.7966
Epoch 43/100
24/24 [==============================] - 14s 594ms/step - loss: 38.7304 - mae: 4.9683 - coeff_determination: 0.7596
Epoch 44/100
24/24 [==============================] - 14s 585ms/step - loss: 34.6333 - mae: 4.6569 - coeff_determination: 0.7868
Epoch 45/100
24/24 [==============================] - 14s 584ms/step - loss: 36.1751 - mae: 4.7532 - coeff_determination: 0.7768
Epoch 46/100
24/24 [==============================] - 14s 591ms/step - loss: 38.5431 - mae: 4.9451 - coeff_determination: 0.7618
Epoch 47/100
24/24 [==============================] - 14s 600ms/step - loss: 39.6571 - mae: 4.9664 - coeff_determination: 0.7563
Epoch 48/100
24/24 [==============================] - 15s 605ms/step - loss: 35.7118 - mae: 4.8633 - coeff_determination: 0.7707
Epoch 49/100
24/24 [==============================] - 15s 607ms/step - loss: 45.5627 - mae: 5.3611 - coeff_determination: 0.7221
Epoch 50/100
24/24 [==============================] - 14s 598ms/step - loss: 32.5951 - mae: 4.5271 - coeff_determination: 0.7984
Epoch 51/100
24/24 [==============================] - 14s 585ms/step - loss: 64.4558 - mae: 6.4162 - coeff_determination: 0.6070
Epoch 52/100
24/24 [==============================] - 14s 598ms/step - loss: 57.5130 - mae: 5.9636 - coeff_determination: 0.6538
Epoch 53/100
24/24 [==============================] - 14s 586ms/step - loss: 33.9630 - mae: 4.6141 - coeff_determination: 0.7920
Epoch 54/100
24/24 [==============================] - 14s 584ms/step - loss: 31.0774 - mae: 4.3929 - coeff_determination: 0.8083
Epoch 55/100
24/24 [==============================] - 14s 588ms/step - loss: 32.2660 - mae: 4.4983 - coeff_determination: 0.8014
Epoch 56/100
24/24 [==============================] - 14s 592ms/step - loss: 33.1396 - mae: 4.5564 - coeff_determination: 0.7953
Epoch 57/100
24/24 [==============================] - 14s 591ms/step - loss: 31.2062 - mae: 4.4033 - coeff_determination: 0.8089
Epoch 58/100
24/24 [==============================] - 14s 594ms/step - loss: 33.0961 - mae: 4.5323 - coeff_determination: 0.7978
Epoch 59/100
24/24 [==============================] - 15s 624ms/step - loss: 32.2997 - mae: 4.5117 - coeff_determination: 0.7968
Epoch 60/100
24/24 [==============================] - 17s 704ms/step - loss: 35.2523 - mae: 4.7187 - coeff_determination: 0.7790
Epoch 61/100
24/24 [==============================] - 17s 717ms/step - loss: 41.3406 - mae: 5.0516 - coeff_determination: 0.7480
Epoch 62/100
24/24 [==============================] - 17s 716ms/step - loss: 31.4312 - mae: 4.4444 - coeff_determination: 0.8043
Epoch 63/100
24/24 [==============================] - 18s 755ms/step - loss: 34.8482 - mae: 4.6490 - coeff_determination: 0.7872
Epoch 64/100
24/24 [==============================] - 17s 706ms/step - loss: 29.0550 - mae: 4.2401 - coeff_determination: 0.8208
Epoch 65/100
24/24 [==============================] - 16s 661ms/step - loss: 28.5786 - mae: 4.2407 - coeff_determination: 0.8234
Epoch 66/100
24/24 [==============================] - 16s 652ms/step - loss: 29.4019 - mae: 4.2420 - coeff_determination: 0.8187
Epoch 67/100
24/24 [==============================] - 16s 676ms/step - loss: 30.5329 - mae: 4.3676 - coeff_determination: 0.8106
Epoch 68/100
24/24 [==============================] - 17s 693ms/step - loss: 34.0045 - mae: 4.5901 - coeff_determination: 0.7912
Epoch 69/100
24/24 [==============================] - 16s 675ms/step - loss: 29.6445 - mae: 4.2796 - coeff_determination: 0.8163
Epoch 70/100
24/24 [==============================] - 17s 717ms/step - loss: 38.3720 - mae: 4.8813 - coeff_determination: 0.7676
Epoch 71/100
24/24 [==============================] - 16s 660ms/step - loss: 31.0481 - mae: 4.4172 - coeff_determination: 0.8087
Epoch 72/100
24/24 [==============================] - 17s 700ms/step - loss: 57.8036 - mae: 6.1670 - coeff_determination: 0.6382
Epoch 73/100
24/24 [==============================] - 16s 666ms/step - loss: 37.5641 - mae: 4.8452 - coeff_determination: 0.7713
Epoch 74/100
24/24 [==============================] - 16s 672ms/step - loss: 33.9289 - mae: 4.6206 - coeff_determination: 0.7899
Epoch 75/100
24/24 [==============================] - 17s 699ms/step - loss: 40.3743 - mae: 5.0730 - coeff_determination: 0.7488
Epoch 76/100
24/24 [==============================] - 16s 682ms/step - loss: 33.4562 - mae: 4.5746 - coeff_determination: 0.7960
Epoch 77/100
24/24 [==============================] - 16s 665ms/step - loss: 32.0642 - mae: 4.4667 - coeff_determination: 0.7992
Epoch 78/100
24/24 [==============================] - 16s 651ms/step - loss: 30.0128 - mae: 4.3519 - coeff_determination: 0.8147
Epoch 79/100
24/24 [==============================] - 18s 770ms/step - loss: 50.3979 - mae: 5.6446 - coeff_determination: 0.6947
Epoch 80/100
24/24 [==============================] - 16s 686ms/step - loss: 48.3741 - mae: 5.5776 - coeff_determination: 0.6998
Epoch 81/100
24/24 [==============================] - 16s 653ms/step - loss: 32.9098 - mae: 4.5715 - coeff_determination: 0.7978
Epoch 82/100
24/24 [==============================] - 16s 654ms/step - loss: 26.9584 - mae: 4.1311 - coeff_determination: 0.8328
Epoch 83/100
24/24 [==============================] - 18s 760ms/step - loss: 31.8625 - mae: 4.4463 - coeff_determination: 0.8046
Epoch 84/100
24/24 [==============================] - 17s 724ms/step - loss: 27.7408 - mae: 4.1910 - coeff_determination: 0.8274
Epoch 85/100
24/24 [==============================] - 15s 632ms/step - loss: 31.5194 - mae: 4.4128 - coeff_determination: 0.8064
Epoch 86/100
24/24 [==============================] - 16s 649ms/step - loss: 26.9652 - mae: 4.1043 - coeff_determination: 0.8336
Epoch 87/100
24/24 [==============================] - 18s 750ms/step - loss: 29.9430 - mae: 4.3444 - coeff_determination: 0.8128
Epoch 88/100
24/24 [==============================] - 15s 627ms/step - loss: 30.9870 - mae: 4.3933 - coeff_determination: 0.8098
Epoch 89/100
24/24 [==============================] - 15s 623ms/step - loss: 36.2213 - mae: 4.7989 - coeff_determination: 0.7747
Epoch 90/100
24/24 [==============================] - 15s 627ms/step - loss: 31.7383 - mae: 4.4464 - coeff_determination: 0.8055
Epoch 91/100
24/24 [==============================] - 15s 617ms/step - loss: 31.6661 - mae: 4.4172 - coeff_determination: 0.8050
Epoch 92/100
24/24 [==============================] - 16s 668ms/step - loss: 35.0071 - mae: 4.6617 - coeff_determination: 0.7817
Epoch 93/100
24/24 [==============================] - 19s 807ms/step - loss: 29.0152 - mae: 4.2270 - coeff_determination: 0.8221
Epoch 94/100
24/24 [==============================] - 18s 761ms/step - loss: 27.3161 - mae: 4.0909 - coeff_determination: 0.8333
Epoch 95/100
24/24 [==============================] - 16s 668ms/step - loss: 29.8845 - mae: 4.2939 - coeff_determination: 0.8175
Epoch 96/100
24/24 [==============================] - 16s 665ms/step - loss: 40.9775 - mae: 5.1339 - coeff_determination: 0.7480
Epoch 97/100
24/24 [==============================] - 16s 659ms/step - loss: 39.3695 - mae: 5.0160 - coeff_determination: 0.7585
Epoch 98/100
24/24 [==============================] - 16s 681ms/step - loss: 35.7855 - mae: 4.7236 - coeff_determination: 0.7804
Epoch 99/100
24/24 [==============================] - 16s 676ms/step - loss: 29.4064 - mae: 4.3035 - coeff_determination: 0.8201
Epoch 100/100
24/24 [==============================] - 16s 670ms/step - loss: 31.3741 - mae: 4.4003 - coeff_determination: 0.8085
---------<TEST RESULTS FOR CLIENT client_2 ; USING LOCAL MODEL>-----------
R^2 0.5002819042801693
Mean squared error 81.15254290922971
Mean absolute error 6.950336584946976
Huber loss 12.92009920759122
SCALING FACTOR : 0.5
--------<TEST RESULTS AFTER ROUND 3 ; USING GLOBAL MODEL>---------
R^2 -0.69
Mean squared error 274.474
Mean absolute error 14.613
Huber loss 16.872
==============================================================
==============================================================
---------<STARTING TRAINING FOR ROUND 4>-----------
---------<STARTING TRAINING FOR CLIENT client_2>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 19s 806ms/step - loss: 329.7290 - mae: 13.6651 - coeff_determination: -0.9642
Epoch 2/100
24/24 [==============================] - 16s 664ms/step - loss: 83.1808 - mae: 7.3575 - coeff_determination: 0.4821
Epoch 3/100
24/24 [==============================] - 16s 655ms/step - loss: 69.2709 - mae: 6.6262 - coeff_determination: 0.5771
Epoch 4/100
24/24 [==============================] - 16s 669ms/step - loss: 54.4393 - mae: 5.8700 - coeff_determination: 0.6654
Epoch 5/100
24/24 [==============================] - 19s 792ms/step - loss: 46.6744 - mae: 5.4269 - coeff_determination: 0.7149
Epoch 6/100
24/24 [==============================] - 17s 705ms/step - loss: 42.9987 - mae: 5.2176 - coeff_determination: 0.7379
Epoch 7/100
24/24 [==============================] - 20s 813ms/step - loss: 46.7490 - mae: 5.4519 - coeff_determination: 0.7138
Epoch 8/100
24/24 [==============================] - 19s 784ms/step - loss: 40.8657 - mae: 5.0984 - coeff_determination: 0.7478
Epoch 9/100
24/24 [==============================] - 18s 740ms/step - loss: 40.6271 - mae: 5.0541 - coeff_determination: 0.7499
Epoch 10/100
24/24 [==============================] - 20s 813ms/step - loss: 41.6847 - mae: 5.1044 - coeff_determination: 0.7450
Epoch 11/100
24/24 [==============================] - 21s 860ms/step - loss: 37.0312 - mae: 4.8106 - coeff_determination: 0.7720
Epoch 12/100
24/24 [==============================] - 18s 745ms/step - loss: 37.2120 - mae: 4.8306 - coeff_determination: 0.7718
Epoch 13/100
24/24 [==============================] - 16s 677ms/step - loss: 48.0432 - mae: 5.4923 - coeff_determination: 0.7076
Epoch 14/100
24/24 [==============================] - 15s 634ms/step - loss: 41.4519 - mae: 5.0602 - coeff_determination: 0.7483
Epoch 15/100
24/24 [==============================] - 16s 680ms/step - loss: 37.8686 - mae: 4.8532 - coeff_determination: 0.7676
Epoch 16/100
24/24 [==============================] - 20s 850ms/step - loss: 45.5534 - mae: 5.4091 - coeff_determination: 0.7099
Epoch 17/100
24/24 [==============================] - 22s 908ms/step - loss: 49.5060 - mae: 5.5436 - coeff_determination: 0.6977
Epoch 18/100
24/24 [==============================] - 19s 783ms/step - loss: 37.8406 - mae: 4.9530 - coeff_determination: 0.7608
Epoch 19/100
24/24 [==============================] - 17s 717ms/step - loss: 42.2553 - mae: 5.1224 - coeff_determination: 0.7428
Epoch 20/100
24/24 [==============================] - 16s 682ms/step - loss: 37.4627 - mae: 4.8265 - coeff_determination: 0.7722
Epoch 21/100
24/24 [==============================] - 16s 667ms/step - loss: 33.1315 - mae: 4.5278 - coeff_determination: 0.7963
Epoch 22/100
24/24 [==============================] - 18s 750ms/step - loss: 39.9652 - mae: 5.0043 - coeff_determination: 0.7558
Epoch 23/100
24/24 [==============================] - 16s 684ms/step - loss: 34.4822 - mae: 4.6549 - coeff_determination: 0.7883
Epoch 24/100
24/24 [==============================] - 19s 774ms/step - loss: 31.4696 - mae: 4.4209 - coeff_determination: 0.8054
Epoch 25/100
24/24 [==============================] - 17s 721ms/step - loss: 30.4774 - mae: 4.3906 - coeff_determination: 0.8100
Epoch 26/100
24/24 [==============================] - 16s 680ms/step - loss: 46.6306 - mae: 5.4381 - coeff_determination: 0.7131
Epoch 27/100
24/24 [==============================] - 16s 663ms/step - loss: 37.1984 - mae: 4.8595 - coeff_determination: 0.7682
Epoch 28/100
24/24 [==============================] - 16s 656ms/step - loss: 34.9714 - mae: 4.6747 - coeff_determination: 0.7849
Epoch 29/100
24/24 [==============================] - 16s 668ms/step - loss: 31.5821 - mae: 4.4671 - coeff_determination: 0.8049
Epoch 30/100
24/24 [==============================] - 18s 749ms/step - loss: 32.2119 - mae: 4.4525 - coeff_determination: 0.8017
Epoch 31/100
24/24 [==============================] - 19s 782ms/step - loss: 32.5023 - mae: 4.5349 - coeff_determination: 0.7978
Epoch 32/100
24/24 [==============================] - 17s 704ms/step - loss: 40.6027 - mae: 5.0236 - coeff_determination: 0.7506
Epoch 33/100
24/24 [==============================] - 17s 699ms/step - loss: 50.8656 - mae: 5.6579 - coeff_determination: 0.6905
Epoch 34/100
24/24 [==============================] - 17s 692ms/step - loss: 42.1978 - mae: 5.1377 - coeff_determination: 0.7381
Epoch 35/100
24/24 [==============================] - 17s 704ms/step - loss: 31.3396 - mae: 4.4228 - coeff_determination: 0.8086
Epoch 36/100
24/24 [==============================] - 17s 702ms/step - loss: 35.6828 - mae: 4.7280 - coeff_determination: 0.7810
Epoch 37/100
24/24 [==============================] - 17s 708ms/step - loss: 40.1741 - mae: 5.0382 - coeff_determination: 0.7528
Epoch 38/100
24/24 [==============================] - 17s 694ms/step - loss: 32.7423 - mae: 4.5179 - coeff_determination: 0.7984
Epoch 39/100
24/24 [==============================] - 17s 699ms/step - loss: 35.9244 - mae: 4.7406 - coeff_determination: 0.7790
Epoch 40/100
24/24 [==============================] - 17s 714ms/step - loss: 28.5643 - mae: 4.2108 - coeff_determination: 0.8244
Epoch 41/100
24/24 [==============================] - 18s 757ms/step - loss: 29.9133 - mae: 4.3215 - coeff_determination: 0.8170
Epoch 42/100
24/24 [==============================] - 16s 684ms/step - loss: 28.4265 - mae: 4.2175 - coeff_determination: 0.8250
Epoch 43/100
24/24 [==============================] - 17s 724ms/step - loss: 29.5041 - mae: 4.2522 - coeff_determination: 0.8192
Epoch 44/100
24/24 [==============================] - 16s 660ms/step - loss: 29.4678 - mae: 4.2509 - coeff_determination: 0.8202
Epoch 45/100
24/24 [==============================] - 16s 656ms/step - loss: 32.9390 - mae: 4.5337 - coeff_determination: 0.7994
Epoch 46/100
24/24 [==============================] - 20s 853ms/step - loss: 41.8966 - mae: 5.1812 - coeff_determination: 0.7394
Epoch 47/100
24/24 [==============================] - 19s 776ms/step - loss: 52.9660 - mae: 5.7268 - coeff_determination: 0.6763
Epoch 48/100
24/24 [==============================] - 15s 640ms/step - loss: 46.5214 - mae: 5.3592 - coeff_determination: 0.7169
Epoch 49/100
24/24 [==============================] - 16s 656ms/step - loss: 36.0353 - mae: 4.7809 - coeff_determination: 0.7774
Epoch 50/100
24/24 [==============================] - 15s 638ms/step - loss: 33.6252 - mae: 4.5684 - coeff_determination: 0.7953
Epoch 51/100
24/24 [==============================] - 16s 661ms/step - loss: 36.0339 - mae: 4.7634 - coeff_determination: 0.7780
Epoch 52/100
24/24 [==============================] - 16s 657ms/step - loss: 47.5640 - mae: 5.5322 - coeff_determination: 0.7066
Epoch 53/100
24/24 [==============================] - 16s 656ms/step - loss: 39.1250 - mae: 4.9503 - coeff_determination: 0.7585
Epoch 54/100
24/24 [==============================] - 16s 661ms/step - loss: 30.1963 - mae: 4.3167 - coeff_determination: 0.8134
Epoch 55/100
24/24 [==============================] - 16s 656ms/step - loss: 26.9871 - mae: 4.1112 - coeff_determination: 0.8323
Epoch 56/100
24/24 [==============================] - 15s 645ms/step - loss: 26.4923 - mae: 4.0700 - coeff_determination: 0.8355
Epoch 57/100
24/24 [==============================] - 16s 646ms/step - loss: 32.1471 - mae: 4.4993 - coeff_determination: 0.8021
Epoch 58/100
24/24 [==============================] - 16s 655ms/step - loss: 34.8131 - mae: 4.6617 - coeff_determination: 0.7865
Epoch 59/100
24/24 [==============================] - 16s 663ms/step - loss: 32.2503 - mae: 4.4575 - coeff_determination: 0.8017
Epoch 60/100
24/24 [==============================] - 16s 659ms/step - loss: 32.4689 - mae: 4.5188 - coeff_determination: 0.7983
Epoch 61/100
24/24 [==============================] - 15s 639ms/step - loss: 30.5120 - mae: 4.3994 - coeff_determination: 0.8065
Epoch 62/100
24/24 [==============================] - 15s 643ms/step - loss: 31.4003 - mae: 4.4199 - coeff_determination: 0.8063
Epoch 63/100
24/24 [==============================] - 17s 698ms/step - loss: 28.7524 - mae: 4.2266 - coeff_determination: 0.8219
Epoch 64/100
24/24 [==============================] - 19s 808ms/step - loss: 27.6282 - mae: 4.1156 - coeff_determination: 0.8317
Epoch 65/100
24/24 [==============================] - 17s 727ms/step - loss: 30.8514 - mae: 4.3791 - coeff_determination: 0.8105
Epoch 66/100
24/24 [==============================] - 16s 681ms/step - loss: 27.9000 - mae: 4.1357 - coeff_determination: 0.8297
Epoch 67/100
24/24 [==============================] - 17s 696ms/step - loss: 44.4957 - mae: 5.3701 - coeff_determination: 0.7195
Epoch 68/100
24/24 [==============================] - 17s 714ms/step - loss: 42.3163 - mae: 5.1028 - coeff_determination: 0.7474
Epoch 69/100
24/24 [==============================] - 17s 708ms/step - loss: 34.1595 - mae: 4.6539 - coeff_determination: 0.7851
Epoch 70/100
24/24 [==============================] - 16s 649ms/step - loss: 29.3232 - mae: 4.2687 - coeff_determination: 0.8210
Epoch 71/100
24/24 [==============================] - 16s 652ms/step - loss: 27.6167 - mae: 4.1228 - coeff_determination: 0.8304
Epoch 72/100
24/24 [==============================] - 15s 627ms/step - loss: 29.5548 - mae: 4.2606 - coeff_determination: 0.8180
Epoch 73/100
24/24 [==============================] - 15s 625ms/step - loss: 30.5151 - mae: 4.3798 - coeff_determination: 0.8139
Epoch 74/100
24/24 [==============================] - 15s 628ms/step - loss: 26.5167 - mae: 4.0589 - coeff_determination: 0.8362
Epoch 75/100
24/24 [==============================] - 15s 633ms/step - loss: 27.3609 - mae: 4.1134 - coeff_determination: 0.8321
Epoch 76/100
24/24 [==============================] - 15s 640ms/step - loss: 48.8304 - mae: 5.5913 - coeff_determination: 0.6995
Epoch 77/100
24/24 [==============================] - 15s 638ms/step - loss: 47.7019 - mae: 5.4899 - coeff_determination: 0.7087
Epoch 78/100
24/24 [==============================] - 16s 677ms/step - loss: 33.4930 - mae: 4.5763 - coeff_determination: 0.7926
Epoch 79/100
24/24 [==============================] - 15s 623ms/step - loss: 28.1789 - mae: 4.1635 - coeff_determination: 0.8266
Epoch 80/100
24/24 [==============================] - 16s 661ms/step - loss: 27.8058 - mae: 4.1553 - coeff_determination: 0.8272
Epoch 81/100
24/24 [==============================] - 16s 662ms/step - loss: 27.5766 - mae: 4.1194 - coeff_determination: 0.8317
Epoch 82/100
24/24 [==============================] - 15s 645ms/step - loss: 24.8118 - mae: 3.9091 - coeff_determination: 0.8441
Epoch 83/100
24/24 [==============================] - 15s 642ms/step - loss: 27.8493 - mae: 4.1771 - coeff_determination: 0.8272
Epoch 84/100
24/24 [==============================] - 15s 645ms/step - loss: 35.5864 - mae: 4.7255 - coeff_determination: 0.7764
Epoch 85/100
24/24 [==============================] - 16s 653ms/step - loss: 34.4130 - mae: 4.6365 - coeff_determination: 0.7901
Epoch 86/100
24/24 [==============================] - 16s 647ms/step - loss: 33.4253 - mae: 4.5544 - coeff_determination: 0.7968
Epoch 87/100
24/24 [==============================] - 16s 647ms/step - loss: 38.4344 - mae: 4.8852 - coeff_determination: 0.7652
Epoch 88/100
24/24 [==============================] - 16s 671ms/step - loss: 32.5958 - mae: 4.5399 - coeff_determination: 0.7992
Epoch 89/100
24/24 [==============================] - 16s 668ms/step - loss: 29.0802 - mae: 4.2279 - coeff_determination: 0.8219
Epoch 90/100
24/24 [==============================] - 15s 642ms/step - loss: 30.3486 - mae: 4.3206 - coeff_determination: 0.8156
Epoch 91/100
24/24 [==============================] - 15s 633ms/step - loss: 44.3268 - mae: 5.2965 - coeff_determination: 0.7245
Epoch 92/100
24/24 [==============================] - 16s 671ms/step - loss: 32.5945 - mae: 4.5144 - coeff_determination: 0.8008
Epoch 93/100
24/24 [==============================] - 15s 641ms/step - loss: 28.0697 - mae: 4.1999 - coeff_determination: 0.8250
Epoch 94/100
24/24 [==============================] - 15s 645ms/step - loss: 29.4225 - mae: 4.2962 - coeff_determination: 0.8188
Epoch 95/100
24/24 [==============================] - 15s 635ms/step - loss: 25.6827 - mae: 4.0029 - coeff_determination: 0.8401
Epoch 96/100
24/24 [==============================] - 15s 628ms/step - loss: 36.9919 - mae: 4.7887 - coeff_determination: 0.7724
Epoch 97/100
24/24 [==============================] - 15s 618ms/step - loss: 28.4039 - mae: 4.2527 - coeff_determination: 0.8246
Epoch 98/100
24/24 [==============================] - 16s 647ms/step - loss: 27.9236 - mae: 4.1662 - coeff_determination: 0.8270
Epoch 99/100
24/24 [==============================] - 16s 649ms/step - loss: 27.7571 - mae: 4.1464 - coeff_determination: 0.8298
Epoch 100/100
24/24 [==============================] - 15s 643ms/step - loss: 27.3318 - mae: 4.1514 - coeff_determination: 0.8314
---------<TEST RESULTS FOR CLIENT client_2 ; USING LOCAL MODEL>-----------
R^2 0.5216889878206556
Mean squared error 77.67610433223894
Mean absolute error 6.7768342822023
Huber loss 12.773990798312507
SCALING FACTOR : 0.5
---------<STARTING TRAINING FOR CLIENT client_1>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
24/24 [==============================] - 22s 900ms/step - loss: 387.6057 - mae: 15.0622 - coeff_determination: -1.3799
Epoch 2/100
24/24 [==============================] - 15s 639ms/step - loss: 90.8386 - mae: 7.6992 - coeff_determination: 0.4325
Epoch 3/100
24/24 [==============================] - 15s 634ms/step - loss: 79.7585 - mae: 7.1408 - coeff_determination: 0.5142
Epoch 4/100
24/24 [==============================] - 15s 634ms/step - loss: 64.1892 - mae: 6.3698 - coeff_determination: 0.6072
Epoch 5/100
24/24 [==============================] - 15s 633ms/step - loss: 58.2567 - mae: 6.0641 - coeff_determination: 0.6438
Epoch 6/100
24/24 [==============================] - 17s 719ms/step - loss: 53.4009 - mae: 5.8216 - coeff_determination: 0.6717
Epoch 7/100
24/24 [==============================] - 24s 1s/step - loss: 45.3709 - mae: 5.3438 - coeff_determination: 0.7195
Epoch 8/100
24/24 [==============================] - 19s 776ms/step - loss: 45.8697 - mae: 5.3866 - coeff_determination: 0.7160
Epoch 9/100
24/24 [==============================] - 21s 870ms/step - loss: 49.9530 - mae: 5.6231 - coeff_determination: 0.6929
Epoch 10/100
24/24 [==============================] - 27s 1s/step - loss: 43.5048 - mae: 5.2681 - coeff_determination: 0.7339
Epoch 11/100
24/24 [==============================] - 17s 706ms/step - loss: 38.3123 - mae: 4.9361 - coeff_determination: 0.7627
Epoch 12/100
24/24 [==============================] - 18s 751ms/step - loss: 41.3056 - mae: 5.1105 - coeff_determination: 0.7480
Epoch 13/100
24/24 [==============================] - 20s 825ms/step - loss: 38.2514 - mae: 4.9236 - coeff_determination: 0.7650
Epoch 14/100
24/24 [==============================] - 21s 880ms/step - loss: 38.9949 - mae: 4.9339 - coeff_determination: 0.7612
Epoch 15/100
24/24 [==============================] - 18s 747ms/step - loss: 35.8292 - mae: 4.7352 - coeff_determination: 0.7772
Epoch 16/100
24/24 [==============================] - 16s 686ms/step - loss: 38.0581 - mae: 4.9410 - coeff_determination: 0.7638
Epoch 17/100
24/24 [==============================] - 16s 664ms/step - loss: 42.1898 - mae: 5.1783 - coeff_determination: 0.7430
Epoch 18/100
24/24 [==============================] - 19s 794ms/step - loss: 37.8145 - mae: 4.8660 - coeff_determination: 0.7673
Epoch 19/100
24/24 [==============================] - 23s 974ms/step - loss: 34.9506 - mae: 4.6564 - coeff_determination: 0.7874
Epoch 20/100
24/24 [==============================] - 19s 802ms/step - loss: 32.1934 - mae: 4.4683 - coeff_determination: 0.8023
Epoch 21/100
24/24 [==============================] - 18s 755ms/step - loss: 34.3137 - mae: 4.6419 - coeff_determination: 0.7869
Epoch 22/100
24/24 [==============================] - 15s 637ms/step - loss: 36.6250 - mae: 4.7988 - coeff_determination: 0.7768
Epoch 23/100
24/24 [==============================] - 18s 759ms/step - loss: 36.5360 - mae: 4.8217 - coeff_determination: 0.7748
Epoch 24/100
24/24 [==============================] - 14s 593ms/step - loss: 33.2754 - mae: 4.5226 - coeff_determination: 0.7954
Epoch 25/100
24/24 [==============================] - 13s 555ms/step - loss: 34.1846 - mae: 4.6120 - coeff_determination: 0.7886
Epoch 26/100
24/24 [==============================] - 14s 580ms/step - loss: 39.4150 - mae: 5.0148 - coeff_determination: 0.7501
Epoch 27/100
24/24 [==============================] - 13s 560ms/step - loss: 39.9669 - mae: 4.9570 - coeff_determination: 0.7579
Epoch 28/100
24/24 [==============================] - 13s 559ms/step - loss: 32.3327 - mae: 4.4943 - coeff_determination: 0.8003
Epoch 29/100
24/24 [==============================] - 13s 551ms/step - loss: 30.3536 - mae: 4.3397 - coeff_determination: 0.8120
Epoch 30/100
24/24 [==============================] - 14s 572ms/step - loss: 30.9421 - mae: 4.3920 - coeff_determination: 0.8068
Epoch 31/100
24/24 [==============================] - 13s 561ms/step - loss: 32.0233 - mae: 4.5001 - coeff_determination: 0.7979
Epoch 32/100
24/24 [==============================] - 13s 555ms/step - loss: 44.8381 - mae: 5.3959 - coeff_determination: 0.7168
Epoch 33/100
24/24 [==============================] - 13s 546ms/step - loss: 35.4420 - mae: 4.6663 - coeff_determination: 0.7837
Epoch 34/100
24/24 [==============================] - 14s 563ms/step - loss: 31.8304 - mae: 4.4233 - coeff_determination: 0.8037
Epoch 35/100
24/24 [==============================] - 13s 554ms/step - loss: 30.2261 - mae: 4.3446 - coeff_determination: 0.8146
Epoch 36/100
24/24 [==============================] - 13s 556ms/step - loss: 31.6756 - mae: 4.4645 - coeff_determination: 0.7996
Epoch 37/100
24/24 [==============================] - 13s 553ms/step - loss: 37.9998 - mae: 4.8819 - coeff_determination: 0.7674
Epoch 38/100
24/24 [==============================] - 13s 550ms/step - loss: 30.5540 - mae: 4.3800 - coeff_determination: 0.8121
Epoch 39/100
24/24 [==============================] - 13s 558ms/step - loss: 35.6594 - mae: 4.7659 - coeff_determination: 0.7817
Epoch 40/100
24/24 [==============================] - 13s 555ms/step - loss: 38.3286 - mae: 4.9214 - coeff_determination: 0.7643
Epoch 41/100
24/24 [==============================] - 13s 557ms/step - loss: 36.9202 - mae: 4.8143 - coeff_determination: 0.7724
Epoch 42/100
24/24 [==============================] - 13s 551ms/step - loss: 39.3939 - mae: 4.9873 - coeff_determination: 0.7585
Epoch 43/100
24/24 [==============================] - 13s 556ms/step - loss: 31.9431 - mae: 4.4667 - coeff_determination: 0.8029
Epoch 44/100
24/24 [==============================] - 13s 556ms/step - loss: 30.7084 - mae: 4.3986 - coeff_determination: 0.8106
Epoch 45/100
24/24 [==============================] - 15s 639ms/step - loss: 30.6171 - mae: 4.3575 - coeff_determination: 0.8111
Epoch 46/100
24/24 [==============================] - 16s 652ms/step - loss: 27.5526 - mae: 4.1394 - coeff_determination: 0.8298
Epoch 47/100
24/24 [==============================] - 19s 808ms/step - loss: 38.7964 - mae: 4.9045 - coeff_determination: 0.7591
Epoch 48/100
24/24 [==============================] - 16s 677ms/step - loss: 38.2821 - mae: 4.9153 - coeff_determination: 0.7641
Epoch 49/100
24/24 [==============================] - 16s 683ms/step - loss: 34.2721 - mae: 4.6187 - coeff_determination: 0.7913
Epoch 50/100
24/24 [==============================] - 15s 627ms/step - loss: 30.3697 - mae: 4.3226 - coeff_determination: 0.8138
Epoch 51/100
24/24 [==============================] - 15s 613ms/step - loss: 30.1495 - mae: 4.3627 - coeff_determination: 0.8122
Epoch 52/100
24/24 [==============================] - 17s 705ms/step - loss: 32.1353 - mae: 4.4780 - coeff_determination: 0.8016
Epoch 53/100
24/24 [==============================] - 15s 624ms/step - loss: 31.9771 - mae: 4.4286 - coeff_determination: 0.8042
Epoch 54/100
24/24 [==============================] - 13s 547ms/step - loss: 30.2335 - mae: 4.3410 - coeff_determination: 0.8138
Epoch 55/100
24/24 [==============================] - 13s 562ms/step - loss: 27.6717 - mae: 4.1754 - coeff_determination: 0.8294
Epoch 56/100
24/24 [==============================] - 14s 567ms/step - loss: 26.3451 - mae: 4.0616 - coeff_determination: 0.8384
Epoch 57/100
24/24 [==============================] - 13s 559ms/step - loss: 29.2632 - mae: 4.2824 - coeff_determination: 0.8196
Epoch 58/100
24/24 [==============================] - 14s 564ms/step - loss: 31.2074 - mae: 4.4088 - coeff_determination: 0.8094
Epoch 59/100
24/24 [==============================] - 22s 918ms/step - loss: 31.3708 - mae: 4.4324 - coeff_determination: 0.8058
Epoch 60/100
24/24 [==============================] - 28s 1s/step - loss: 32.5388 - mae: 4.5590 - coeff_determination: 0.7981
Epoch 61/100
24/24 [==============================] - 28s 1s/step - loss: 36.4840 - mae: 4.7787 - coeff_determination: 0.7772
Epoch 62/100
24/24 [==============================] - 18s 755ms/step - loss: 29.9100 - mae: 4.3095 - coeff_determination: 0.8144
Epoch 63/100
24/24 [==============================] - 21s 894ms/step - loss: 27.4183 - mae: 4.1251 - coeff_determination: 0.8306
Epoch 64/100
24/24 [==============================] - 17s 722ms/step - loss: 30.2739 - mae: 4.3427 - coeff_determination: 0.8134
Epoch 65/100
24/24 [==============================] - 16s 686ms/step - loss: 27.2544 - mae: 4.0860 - coeff_determination: 0.8329
Epoch 66/100
24/24 [==============================] - 15s 638ms/step - loss: 37.2966 - mae: 4.8164 - coeff_determination: 0.7735
Epoch 67/100
24/24 [==============================] - 15s 614ms/step - loss: 32.9273 - mae: 4.5331 - coeff_determination: 0.7985
Epoch 68/100
24/24 [==============================] - 15s 606ms/step - loss: 27.4392 - mae: 4.1196 - coeff_determination: 0.8315
Epoch 69/100
24/24 [==============================] - 15s 614ms/step - loss: 28.7904 - mae: 4.2434 - coeff_determination: 0.8227
Epoch 70/100
24/24 [==============================] - 20s 834ms/step - loss: 27.9593 - mae: 4.1420 - coeff_determination: 0.8264
Epoch 71/100
24/24 [==============================] - 16s 681ms/step - loss: 41.6969 - mae: 5.1071 - coeff_determination: 0.7436
Epoch 72/100
24/24 [==============================] - 15s 617ms/step - loss: 26.9057 - mae: 4.0811 - coeff_determination: 0.8354
Epoch 73/100
24/24 [==============================] - 16s 667ms/step - loss: 25.6561 - mae: 3.9874 - coeff_determination: 0.8427
Epoch 74/100
24/24 [==============================] - 19s 810ms/step - loss: 36.5726 - mae: 4.7702 - coeff_determination: 0.7744
Epoch 75/100
24/24 [==============================] - 19s 782ms/step - loss: 34.3009 - mae: 4.6279 - coeff_determination: 0.7890
Epoch 76/100
24/24 [==============================] - 17s 724ms/step - loss: 35.0261 - mae: 4.7418 - coeff_determination: 0.7824
Epoch 77/100
24/24 [==============================] - 15s 627ms/step - loss: 49.1010 - mae: 5.6796 - coeff_determination: 0.6874
Epoch 78/100
24/24 [==============================] - 15s 625ms/step - loss: 35.3682 - mae: 4.7096 - coeff_determination: 0.7853
Epoch 79/100
24/24 [==============================] - 15s 625ms/step - loss: 29.1051 - mae: 4.2755 - coeff_determination: 0.8188
Epoch 80/100
24/24 [==============================] - 15s 611ms/step - loss: 29.1550 - mae: 4.2807 - coeff_determination: 0.8213
Epoch 81/100
24/24 [==============================] - 15s 633ms/step - loss: 32.8928 - mae: 4.5822 - coeff_determination: 0.7928
Epoch 82/100
24/24 [==============================] - 15s 631ms/step - loss: 32.4686 - mae: 4.4735 - coeff_determination: 0.8011
Epoch 83/100
24/24 [==============================] - 14s 603ms/step - loss: 31.4926 - mae: 4.4531 - coeff_determination: 0.8078
Epoch 84/100
24/24 [==============================] - 15s 632ms/step - loss: 26.3748 - mae: 4.0203 - coeff_determination: 0.8374
Epoch 85/100
24/24 [==============================] - 18s 759ms/step - loss: 27.8290 - mae: 4.1863 - coeff_determination: 0.8270
Epoch 86/100
24/24 [==============================] - 17s 725ms/step - loss: 35.2568 - mae: 4.7544 - coeff_determination: 0.7797
Epoch 87/100
24/24 [==============================] - 17s 720ms/step - loss: 26.2382 - mae: 4.0300 - coeff_determination: 0.8376
Epoch 88/100
24/24 [==============================] - 14s 599ms/step - loss: 27.5494 - mae: 4.1326 - coeff_determination: 0.8302
Epoch 89/100
24/24 [==============================] - 14s 600ms/step - loss: 32.3824 - mae: 4.4646 - coeff_determination: 0.8012
Epoch 90/100
24/24 [==============================] - 14s 588ms/step - loss: 38.3998 - mae: 4.9170 - coeff_determination: 0.7674
Epoch 91/100
24/24 [==============================] - 15s 608ms/step - loss: 35.1872 - mae: 4.6982 - coeff_determination: 0.7860
Epoch 92/100
24/24 [==============================] - 14s 582ms/step - loss: 33.6785 - mae: 4.6399 - coeff_determination: 0.7904
Epoch 93/100
24/24 [==============================] - 14s 595ms/step - loss: 29.9077 - mae: 4.3031 - coeff_determination: 0.8170
Epoch 94/100
24/24 [==============================] - 14s 591ms/step - loss: 25.8628 - mae: 4.0351 - coeff_determination: 0.8381
Epoch 95/100
24/24 [==============================] - 15s 605ms/step - loss: 26.6560 - mae: 4.0504 - coeff_determination: 0.8361
Epoch 96/100
24/24 [==============================] - 14s 597ms/step - loss: 32.6056 - mae: 4.4714 - coeff_determination: 0.8015
Epoch 97/100
24/24 [==============================] - 15s 608ms/step - loss: 29.3722 - mae: 4.2690 - coeff_determination: 0.8206
Epoch 98/100
24/24 [==============================] - 14s 587ms/step - loss: 25.4440 - mae: 3.9829 - coeff_determination: 0.8408
Epoch 99/100
24/24 [==============================] - 14s 595ms/step - loss: 25.0147 - mae: 3.9273 - coeff_determination: 0.8462
Epoch 100/100
24/24 [==============================] - 14s 597ms/step - loss: 25.1098 - mae: 3.9621 - coeff_determination: 0.8448
---------<TEST RESULTS FOR CLIENT client_1 ; USING LOCAL MODEL>-----------
R^2 0.5112753933774674
Mean squared error 79.36723714717698
Mean absolute error 6.75066868821508
Huber loss 12.848073473751517
SCALING FACTOR : 0.5
--------<TEST RESULTS AFTER ROUND 4 ; USING GLOBAL MODEL>---------
R^2 0.235
Mean squared error 124.264
Mean absolute error 9.248
Huber loss 13.452
==============================================================
==============================================================
---------<STARTING TRAINING FOR ROUND 5>-----------
---------<STARTING TRAINING FOR CLIENT client_1>-----------
Model: "model"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_1 (InputLayer) [(None, 18388)] 0
_________________________________________________________________
dense (Dense) (None, 1024) 18830336
_________________________________________________________________
dropout (Dropout) (None, 1024) 0
_________________________________________________________________
dense_1 (Dense) (None, 512) 524800
_________________________________________________________________
dropout_1 (Dropout) (None, 512) 0
_________________________________________________________________
dense_2 (Dense) (None, 64) 32832
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
age_output (Dense) (None, 1) 65
=================================================================
Total params: 19,388,033
Trainable params: 19,388,033
Non-trainable params: 0
_________________________________________________________________
Epoch 1/100
14/Unknown - 11s 792ms/step - loss: 892.6613 - mae: 24.2433 - coeff_determination: -4.5195 |
playbook/tactics/privilege-escalation/T1546.013.ipynb | ###Markdown
T1546.013 - Event Triggered Execution: PowerShell ProfileAdversaries may gain persistence and elevate privileges by executing malicious content triggered by PowerShell profiles. A PowerShell profile (profile.ps1) is a script that runs when [PowerShell](https://attack.mitre.org/techniques/T1059/001) starts and can be used as a logon script to customize user environments.[PowerShell](https://attack.mitre.org/techniques/T1059/001) supports several profiles depending on the user or host program. For example, there can be different profiles for [PowerShell](https://attack.mitre.org/techniques/T1059/001) host programs such as the PowerShell console, PowerShell ISE or Visual Studio Code. An administrator can also configure a profile that applies to all users and host programs on the local computer. (Citation: Microsoft About Profiles) Adversaries may modify these profiles to include arbitrary commands, functions, modules, and/or [PowerShell](https://attack.mitre.org/techniques/T1059/001) drives to gain persistence. Every time a user opens a [PowerShell](https://attack.mitre.org/techniques/T1059/001) session the modified script will be executed unless the -NoProfile flag is used when it is launched. (Citation: ESET Turla PowerShell May 2019) An adversary may also be able to escalate privileges if a script in a PowerShell profile is loaded and executed by an account with higher privileges, such as a domain administrator. (Citation: Wits End and Shady PowerShell Profiles) Atomic Tests
###Code
#Import the Module before running the tests.
# Checkout Jupyter Notebook at https://github.com/cyb3rbuff/TheAtomicPlaybook to run PS scripts.
Import-Module /Users/0x6c/AtomicRedTeam/atomics/invoke-atomicredteam/Invoke-AtomicRedTeam.psd1 - Force
###Output
_____no_output_____
###Markdown
Atomic Test 1 - Append malicious start-process cmdletAppends a start process cmdlet to the current user's powershell profile pofile that points to a malicious executable. Upon execution, calc.exe will be launched.**Supported Platforms:** windows Dependencies: Run with `powershell`! Description: Ensure a powershell profile exists for the current user Check Prereq Commands:```powershellif (Test-Path $profile) {exit 0} else {exit 1}``` Get Prereq Commands:```powershellNew-Item -Path $profile -Type File -Force```
###Code
Invoke-AtomicTest T1546.013 -TestNumbers 1 -GetPreReqs
###Output
_____no_output_____
###Markdown
Attack Commands: Run with `powershell````powershellAdd-Content $profile -Value ""Add-Content $profile -Value "Start-Process calc.exe"powershell -Command exit```
###Code
Invoke-AtomicTest T1546.013 -TestNumbers 1
###Output
_____no_output_____ |
Recommendation Study/Recommender system.ipynb | ###Markdown
추천시스템 (Recommende System)- 추천시스템은 크게 두가지로 구분가능 - 컨텐츠 기반 필터링(content-based filtering) - 협업 필터링(collaborative filtering)- 두가지를 조합한 Hybrid 방식도 가능- 컨텐츠 기반 필터링은 지금까지 사용자의 이전행동과 명시적 피드백을 통해 사용자가 좋아하는 것과 유사한 항목을 추천- 협업 필터링은 사용자와 항목간의 유사성을 동시에 사용해 추천 Surprise- 추천 시스템 개발을 위한 라이브러리- 다양한 모델과 데이터 제공- scikit-learn과 유사한 사용 방법
###Code
from surprise import SVD
from surprise import Dataset
from surprise.model_selection import cross_validate
from sklearn.decomposition import randomized_svd, non_negative_factorization
import numpy as np
data = Dataset.load_builtin('ml-100k',prompt=False)
data.raw_ratings[:10]
#data 구성
#user/item(movie)/score(rating)/ID
model=SVD()
cross_validate(model, data, measures=['rmse', 'mae'], cv=5, verbose=True)
###Output
Evaluating RMSE, MAE of algorithm SVD on 5 split(s).
Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Mean Std
RMSE (testset) 0.9298 0.9357 0.9473 0.9354 0.9308 0.9358 0.0062
MAE (testset) 0.7328 0.7403 0.7476 0.7373 0.7316 0.7379 0.0058
Fit time 4.67 4.76 4.80 4.82 5.00 4.81 0.11
Test time 0.18 0.18 0.16 0.12 0.16 0.16 0.02
###Markdown
컨텐츠 기반 필터링(Content-based Filtering)- 컨텐츠 기반 필터링은 이전의 행동과 명시적 피드백을 통해 좋아하는 것과 유사한 항목을 추천 - 예를들어 내가 지금까지 시청한 영화 목록과 다른 사용자의 시청 목록을 비교해 나와 비슷한 취향의 사용자가 시청한 영화를 추천- 유사도를 기반으로 추천- 컨텐츠 기반 필터링은 다음과 같은 장단점이 있다. - 장점 - 많은 수의 사용자를 대상으로 쉽게 확장이 가능 - 사용자가 관심을 갖지 않던 상품 추천가능 - 단점 - 입력 특성을 직접 설계해야 하기 때문에 많은 도메인 지식이 필요 - 사용자의 기존 관심사항을 기반으로만 추천 가능 - 콜드 스타트 문제 존재
###Code
data = Dataset.load_builtin('ml-100k', prompt=False)
raw_data = np.array(data.raw_ratings, dtype=int)
raw_data
raw_data[:,0]-=1
raw_data[:,1]-=1
raw_data
n_users = np.max(raw_data[:,0])
n_movies = np.max(raw_data[:,1])
shape = (n_users+1, n_movies+1)
shape
adj_matrix = np.ndarray(shape,dtype=int)
for user_id, movie_id, rating, time in raw_data:
adj_matrix[user_id][movie_id]=1.
adj_matrix
my_id, my_vector = 0, adj_matrix[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(adj_matrix):
if my_id != user_id:
similarity = np.dot(my_vector, user_vector)
if similarity > best_match:
best_match = similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, log in enumerate(zip(my_vector, best_match_vector)):
log1, log2 = log
if log1<1. and log2>0.:
recommed_list.append(i)
print(recommed_list)
###Output
[272, 273, 275, 280, 281, 283, 287, 288, 289, 290, 292, 293, 297, 299, 300, 301, 302, 306, 312, 314, 315, 316, 317, 321, 322, 323, 324, 327, 330, 331, 332, 333, 339, 342, 345, 346, 353, 354, 355, 356, 357, 363, 364, 365, 366, 372, 374, 378, 379, 381, 382, 383, 384, 385, 386, 387, 390, 391, 392, 394, 395, 396, 398, 400, 401, 402, 403, 404, 405, 406, 407, 408, 409, 410, 412, 414, 416, 417, 418, 419, 420, 422, 424, 425, 426, 427, 428, 430, 431, 432, 435, 442, 446, 447, 448, 449, 450, 451, 452, 454, 455, 457, 460, 461, 462, 468, 469, 470, 471, 472, 473, 474, 478, 495, 500, 507, 517, 522, 525, 530, 539, 540, 543, 545, 546, 548, 549, 550, 551, 553, 557, 558, 560, 561, 562, 563, 565, 566, 567, 568, 570, 571, 574, 575, 576, 577, 580, 581, 582, 585, 587, 589, 590, 594, 596, 602, 623, 626, 627, 630, 633, 635, 639, 646, 648, 651, 652, 654, 657, 664, 668, 671, 677, 678, 681, 683, 684, 685, 690, 691, 692, 695, 696, 708, 709, 714, 718, 719, 720, 724, 726, 727, 731, 733, 734, 736, 738, 741, 742, 745, 746, 747, 749, 750, 754, 758, 762, 764, 767, 768, 769, 770, 771, 772, 773, 778, 779, 782, 785, 788, 789, 793, 795, 796, 799, 800, 801, 802, 805, 806, 808, 815, 819, 822, 824, 830, 839, 842, 843, 844, 852, 853, 870, 875, 878, 880, 889, 901, 914, 915, 918, 921, 927, 929, 930, 938, 940, 941, 942, 948, 950, 958, 968, 973, 974, 976, 992, 999, 1005, 1009, 1010, 1012, 1015, 1018, 1027, 1030, 1034, 1035, 1041, 1043, 1045, 1046, 1051, 1055, 1072, 1073, 1078, 1080, 1082, 1088, 1089, 1090, 1094, 1097, 1108, 1109, 1117, 1128, 1130, 1134, 1139, 1140, 1144, 1156, 1169, 1171, 1179, 1193, 1198, 1207, 1209, 1212, 1217, 1219, 1220, 1227, 1231, 1238, 1239, 1243, 1244, 1252, 1266, 1272, 1273, 1300, 1313, 1406, 1412, 1415, 1470, 1477, 1480, 1481, 1482]
###Markdown
유클리드 거리를 사용해 추천$$euclidean=\ \sqrt{\sum_{d=1}^{D\ }{(A_i-B_i)^2}} $$- 거리가 가까울 수록(값이 작을 수록) 나와 유사항 사용자
###Code
my_id, my_vector = 0, adj_matrix[0]
best_match, best_match_id, best_match_vector = 999,-1,[]
for user_id, user_vector in enumerate(adj_matrix):
if my_id != user_id:
euclidean_dist = np.sqrt(np.sum(np.square(my_vector - user_vector)))
if euclidean_dist < best_match:
best_match = euclidean_dist
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, log in enumerate(zip(my_vector, best_match_vector)):
log1, log2 = log
if log1<1. and log2>0.:
recommed_list.append(i)
print(recommed_list)
###Output
[297, 312, 317, 342, 356, 366, 379, 384, 392, 402, 404, 407, 417, 422, 428, 433, 448, 454, 469, 473, 495, 510, 516, 526, 527, 549, 567, 602, 635, 649, 650, 654, 658, 661, 664, 696, 731, 746, 750, 754, 915, 918, 925, 929, 950, 968, 1015, 1046]
###Markdown
코사인 유사도를 사용해 추천$$cos\theta = \dfrac{A \bullet B}{\rVert A \rVert \times \rVert B \rVert}$$- 두 벡터가 이루고 있는 각을 계산
###Code
def compute_cos_similarity(v1,v2):
norm1 = np.sqrt(np.sum(np.square(v1)))
norm2 = np.sqrt(np.sum(np.square(v2)))
dot = np.dot(v1,v2)
return dot / (norm1*norm2)
my_id, my_vector = 0, adj_matrix[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(adj_matrix):
if my_id != user_id:
cos_similarity = compute_cos_similarity(my_vector, user_vector)
if cos_similarity > best_match:
best_match = cos_similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, log in enumerate(zip(my_vector, best_match_vector)):
log1, log2 = log
if log1<1. and log2>0.:
recommed_list.append(i)
print(recommed_list)
###Output
[272, 275, 279, 280, 283, 285, 289, 294, 297, 316, 317, 355, 365, 366, 368, 379, 380, 381, 384, 386, 392, 398, 401, 404, 416, 420, 422, 424, 426, 427, 430, 432, 450, 460, 461, 466, 469, 471, 473, 474, 475, 479, 482, 483, 497, 505, 508, 510, 511, 522, 526, 527, 529, 530, 534, 536, 540, 545, 548, 549, 556, 557, 558, 560, 565, 567, 568, 569, 577, 580, 581, 582, 592, 596, 630, 635, 639, 641, 649, 651, 654, 673, 677, 678, 683, 684, 692, 696, 701, 703, 707, 708, 709, 712, 714, 719, 720, 726, 731, 734, 736, 738, 740, 745, 747, 754, 755, 761, 762, 763, 766, 780, 789, 791, 805, 819, 823, 824, 830, 843, 862, 865, 918, 929, 930, 938, 942, 943, 947, 958, 959, 960, 970, 977, 1004, 1008, 1009, 1010, 1013, 1041, 1045, 1069, 1072, 1073, 1078, 1097, 1100, 1108, 1112, 1118, 1134, 1193, 1205, 1207, 1216, 1219, 1267, 1334, 1400, 1427, 1596, 1681]
###Markdown
기존 방법에 명시적 피드백(사용자가 평가한 영화 점수)을 추가해 실험- 위 테스트 까지는 봣는지 안봤는지 0,1로만 구분했지만,- rating 점수를 추가 기입
###Code
adj_matrix = np.ndarray(shape,dtype=int)
for user_id, movie_id, rating, time in raw_data:
adj_matrix[user_id][movie_id] = rating
adj_matrix
my_id, my_vector = 0, adj_matrix[0]
best_match, best_match_id, best_match_vector = 999,-1,[]
for user_id, user_vector in enumerate(adj_matrix):
if my_id != user_id:
euclidean_dist = np.sqrt(np.sum(np.square(my_vector - user_vector)))
if euclidean_dist < best_match:
best_match = euclidean_dist
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
my_id, my_vector = 0, adj_matrix[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(adj_matrix):
if my_id != user_id:
cos_similarity = compute_cos_similarity(my_vector, user_vector)
if cos_similarity > best_match:
best_match = cos_similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
###Output
Best Match : 0.569065731527988, Best Match ID : 915
###Markdown
협업 필터링(Collaborative Filtering)- 사용자가 항목의 유사성을 동시에 고려해 추천- 기존에 내 관심사가 아닌 항목이라도 추천 가능- 자동으로 임베딩 학습 가능- 협업 필터링은 다음과 같은 장단점을 갖고 있다. - 장점 - 자동으로 임베딜을 학습하기 때문에 도메인 지식이 필요 없다. - 기존의 관심사가 아니더라도 추천 가능 - 단점 - 학습 과정에 나오지 않은 항목은 임베딩을 만들 수 없음 - 추가 특성을 사용하기 어려움
###Code
from surprise import KNNBasic, SVD, SVDpp, NMF
from surprise.model_selection import cross_validate
data = Dataset.load_builtin('ml-100k',prompt=False)
###Output
_____no_output_____
###Markdown
KNN
###Code
model = KNNBasic()
cross_validate(model, data, measures=['rmse','mae'],cv=5,n_jobs=4,verbose=True)
###Output
Evaluating RMSE, MAE of algorithm KNNBasic on 5 split(s).
Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Mean Std
RMSE (testset) 0.9801 0.9761 0.9770 0.9886 0.9750 0.9794 0.0049
MAE (testset) 0.7739 0.7722 0.7681 0.7825 0.7693 0.7732 0.0051
Fit time 0.23 0.29 0.36 0.34 0.35 0.31 0.05
Test time 3.73 4.37 4.57 3.81 3.00 3.90 0.55
###Markdown
SVD
###Code
model = SVD()
cross_validate(model, data, measures=['rmse','mae'],cv=5,n_jobs=4,verbose=True)
###Output
Evaluating RMSE, MAE of algorithm SVD on 5 split(s).
Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Mean Std
RMSE (testset) 0.9312 0.9389 0.9403 0.9316 0.9365 0.9357 0.0037
MAE (testset) 0.7352 0.7429 0.7385 0.7368 0.7361 0.7379 0.0027
Fit time 8.09 9.47 9.21 7.99 6.38 8.23 1.10
Test time 0.31 0.28 0.21 0.14 0.14 0.22 0.07
###Markdown
NMF
###Code
model = NMF()
cross_validate(model, data, measures=['rmse','mae'],cv=5,n_jobs=4,verbose=True)
###Output
Evaluating RMSE, MAE of algorithm NMF on 5 split(s).
Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Mean Std
RMSE (testset) 0.9696 0.9539 0.9657 0.9569 0.9675 0.9627 0.0062
MAE (testset) 0.7622 0.7496 0.7594 0.7522 0.7606 0.7568 0.0050
Fit time 6.31 7.91 7.90 6.73 5.21 6.81 1.02
Test time 0.31 0.19 0.17 0.11 0.12 0.18 0.07
###Markdown
SVD++ (need long time)
###Code
model = SVDpp()
cross_validate(model, data, measures=['rmse','mae'],cv=5,n_jobs=4,verbose=True)
###Output
Evaluating RMSE, MAE of algorithm SVDpp on 5 split(s).
Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Mean Std
RMSE (testset) 0.9176 0.9236 0.9195 0.9160 0.9105 0.9174 0.0043
MAE (testset) 0.7186 0.7234 0.7213 0.7197 0.7129 0.7192 0.0035
Fit time 331.22 332.18 331.31 333.60 158.24 297.31 69.54
Test time 5.16 5.26 5.00 2.82 2.42 4.13 1.24
###Markdown
하이브리드(Hybrid)- 컨텐츠 기반 필터링과 협업 필터링을 조합한 방식- 많은 하이브리드 방식이 존재- 실습에서는 협업 필터링으로 임베딩을 학습하고 컨텐츠 기반 필터링으로 유사도 기반 추천을 수행하는 추천 엔진 개발
###Code
data = Dataset.load_builtin('ml-100k',prompt=False)
raw_data = np.array(data.raw_ratings, dtype=int)
raw_data[:,0]-=1
raw_data[:,1]-=1
n_users = np.max(raw_data[:,0])
n_movies = np.max(raw_data[:,1])
shape = (n_users+1,n_movies+1)
shape
adj_matrix = np.ndarray(shape, dtype=int)
for user_id, movie_id, rating, time in raw_data:
adj_matrix[user_id][movie_id] = rating
adj_matrix
###Output
_____no_output_____
###Markdown
- 행렬 분해
###Code
U, S, V = randomized_svd(adj_matrix, n_components=2)
S = np.diag(S)
print(U.shape)
print(S.shape)
print(V.shape)
np.matmul(np.matmul(U,S),V)
###Output
_____no_output_____
###Markdown
- 사용자 기반 추천- 나와 비슷한 취향을 가진 다른 사용자의 행동을 추천- 사용자 특징 벡터의 유사도 사용
###Code
my_id, my_vector = 0, U[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(U):
if my_id != user_id:
cos_similarity = compute_cos_similarity(my_vector, user_vector)
if cos_similarity > best_match:
best_match = cos_similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, log in enumerate(zip(adj_matrix[my_id], adj_matrix[best_match_id])):
log1, log2 = log
if log1<1. and log2>0.:
recommed_list.append(i)
print(recommed_list)
###Output
[272, 273, 274, 281, 285, 288, 293, 297, 303, 306, 312, 317, 327, 332, 369, 410, 418, 419, 422, 426, 428, 431, 434, 442, 461, 475, 477, 482, 495, 503, 504, 505, 506, 509, 519, 520, 522, 525, 531, 545, 548, 590, 594, 595, 613, 631, 654, 658, 660, 672, 684, 685, 691, 695, 698, 704, 716, 728, 734, 749, 755, 863, 865, 933, 1012, 1038, 1101, 1327, 1400]
###Markdown
- 항목 기반 추천- 내가 본 항목과 비슷한 항목을 추천- 항목 특징 벡터의 유사도 사용
###Code
my_id, my_vector = 0, V.T[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(V.T):
if my_id != user_id:
cos_similarity = compute_cos_similarity(my_vector, user_vector)
if cos_similarity > best_match:
best_match = cos_similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, user_vector in enumerate(adj_matrix):
if adj_matrix[i][my_id] > 0.9:
recommed_list.append(i)
print(recommed_list)
###Output
[0, 1, 4, 5, 9, 12, 14, 15, 16, 17, 19, 20, 22, 24, 25, 37, 40, 41, 42, 43, 44, 48, 53, 55, 56, 57, 58, 61, 62, 63, 64, 65, 66, 69, 71, 72, 74, 76, 78, 80, 81, 82, 83, 88, 91, 92, 93, 94, 95, 96, 98, 100, 101, 105, 107, 108, 116, 119, 120, 123, 124, 127, 129, 130, 133, 136, 137, 140, 143, 144, 147, 149, 150, 156, 157, 159, 161, 167, 173, 176, 177, 180, 181, 183, 188, 192, 193, 197, 198, 199, 200, 201, 202, 203, 208, 209, 212, 215, 221, 222, 229, 230, 231, 233, 234, 241, 242, 243, 245, 246, 247, 248, 249, 250, 251, 252, 253, 255, 261, 262, 264, 267, 270, 273, 274, 275, 276, 278, 279, 285, 286, 288, 289, 290, 291, 292, 293, 294, 295, 296, 297, 298, 300, 302, 304, 306, 307, 310, 311, 312, 313, 319, 321, 323, 324, 325, 326, 329, 330, 331, 335, 337, 338, 339, 342, 343, 344, 346, 347, 349, 356, 358, 359, 362, 364, 370, 373, 377, 378, 379, 380, 386, 387, 388, 389, 392, 393, 394, 395, 397, 398, 400, 401, 402, 405, 406, 410, 411, 415, 416, 418, 421, 423, 424, 428, 431, 433, 434, 437, 440, 444, 446, 449, 453, 454, 455, 456, 457, 458, 459, 462, 464, 466, 467, 469, 470, 471, 477, 478, 482, 483, 485, 486, 487, 489, 492, 493, 494, 496, 499, 502, 504, 507, 511, 513, 516, 517, 520, 522, 524, 525, 531, 532, 533, 534, 535, 536, 539, 540, 541, 544, 547, 548, 549, 551, 552, 553, 559, 560, 561, 566, 568, 575, 576, 578, 579, 581, 587, 591, 592, 596, 598, 601, 604, 605, 608, 609, 611, 612, 613, 617, 619, 620, 621, 623, 629, 631, 633, 634, 635, 636, 641, 642, 647, 648, 649, 652, 653, 654, 656, 657, 659, 660, 662, 663, 664, 668, 673, 675, 676, 677, 678, 679, 681, 683, 688, 689, 690, 691, 696, 697, 698, 700, 702, 704, 705, 707, 708, 709, 713, 714, 715, 720, 722, 725, 726, 729, 730, 732, 734, 737, 741, 743, 744, 745, 746, 747, 748, 750, 755, 756, 758, 760, 762, 763, 766, 767, 768, 769, 770, 772, 776, 778, 784, 785, 787, 788, 789, 791, 792, 793, 794, 795, 797, 799, 803, 804, 805, 806, 814, 816, 820, 821, 822, 825, 828, 829, 830, 834, 837, 838, 842, 846, 851, 853, 863, 864, 866, 867, 869, 871, 878, 879, 880, 881, 882, 884, 885, 886, 888, 889, 891, 892, 893, 894, 895, 896, 898, 900, 901, 902, 906, 909, 912, 915, 916, 917, 918, 920, 921, 922, 923, 926, 928, 929, 931, 932, 933, 934, 935, 937, 940]
###Markdown
- 비음수 행렬 분해를 사용한 하이브리드 추천
###Code
adj_matrix
A, B, iter = non_negative_factorization(adj_matrix,n_components=2)
np.matmul(A,B)
###Output
_____no_output_____
###Markdown
- 사용자 기반 추천
###Code
my_id, my_vector = 0, U[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(U):
if my_id != user_id:
cos_similarity = compute_cos_similarity(my_vector, user_vector)
if cos_similarity > best_match:
best_match = cos_similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, log in enumerate(zip(adj_matrix[my_id], adj_matrix[best_match_id])):
log1, log2 = log
if log1<1. and log2>0.:
recommed_list.append(i)
print(recommed_list)
###Output
[272, 273, 274, 281, 285, 288, 293, 297, 303, 306, 312, 317, 327, 332, 369, 410, 418, 419, 422, 426, 428, 431, 434, 442, 461, 475, 477, 482, 495, 503, 504, 505, 506, 509, 519, 520, 522, 525, 531, 545, 548, 590, 594, 595, 613, 631, 654, 658, 660, 672, 684, 685, 691, 695, 698, 704, 716, 728, 734, 749, 755, 863, 865, 933, 1012, 1038, 1101, 1327, 1400]
###Markdown
- 항목 기반 추천
###Code
my_id, my_vector = 0, V.T[0]
best_match, best_match_id, best_match_vector = -1,-1,[]
for user_id, user_vector in enumerate(V.T):
if my_id != user_id:
cos_similarity = compute_cos_similarity(my_vector, user_vector)
if cos_similarity > best_match:
best_match = cos_similarity
best_match_id = user_id
best_match_vector = user_vector
print('Best Match : {}, Best Match ID : {}'.format(best_match,best_match_id))
recommed_list = []
for i, user_vector in enumerate(adj_matrix):
if adj_matrix[i][my_id] > 0.9:
recommed_list.append(i)
print(recommed_list)
###Output
[0, 1, 4, 5, 9, 12, 14, 15, 16, 17, 19, 20, 22, 24, 25, 37, 40, 41, 42, 43, 44, 48, 53, 55, 56, 57, 58, 61, 62, 63, 64, 65, 66, 69, 71, 72, 74, 76, 78, 80, 81, 82, 83, 88, 91, 92, 93, 94, 95, 96, 98, 100, 101, 105, 107, 108, 116, 119, 120, 123, 124, 127, 129, 130, 133, 136, 137, 140, 143, 144, 147, 149, 150, 156, 157, 159, 161, 167, 173, 176, 177, 180, 181, 183, 188, 192, 193, 197, 198, 199, 200, 201, 202, 203, 208, 209, 212, 215, 221, 222, 229, 230, 231, 233, 234, 241, 242, 243, 245, 246, 247, 248, 249, 250, 251, 252, 253, 255, 261, 262, 264, 267, 270, 273, 274, 275, 276, 278, 279, 285, 286, 288, 289, 290, 291, 292, 293, 294, 295, 296, 297, 298, 300, 302, 304, 306, 307, 310, 311, 312, 313, 319, 321, 323, 324, 325, 326, 329, 330, 331, 335, 337, 338, 339, 342, 343, 344, 346, 347, 349, 356, 358, 359, 362, 364, 370, 373, 377, 378, 379, 380, 386, 387, 388, 389, 392, 393, 394, 395, 397, 398, 400, 401, 402, 405, 406, 410, 411, 415, 416, 418, 421, 423, 424, 428, 431, 433, 434, 437, 440, 444, 446, 449, 453, 454, 455, 456, 457, 458, 459, 462, 464, 466, 467, 469, 470, 471, 477, 478, 482, 483, 485, 486, 487, 489, 492, 493, 494, 496, 499, 502, 504, 507, 511, 513, 516, 517, 520, 522, 524, 525, 531, 532, 533, 534, 535, 536, 539, 540, 541, 544, 547, 548, 549, 551, 552, 553, 559, 560, 561, 566, 568, 575, 576, 578, 579, 581, 587, 591, 592, 596, 598, 601, 604, 605, 608, 609, 611, 612, 613, 617, 619, 620, 621, 623, 629, 631, 633, 634, 635, 636, 641, 642, 647, 648, 649, 652, 653, 654, 656, 657, 659, 660, 662, 663, 664, 668, 673, 675, 676, 677, 678, 679, 681, 683, 688, 689, 690, 691, 696, 697, 698, 700, 702, 704, 705, 707, 708, 709, 713, 714, 715, 720, 722, 725, 726, 729, 730, 732, 734, 737, 741, 743, 744, 745, 746, 747, 748, 750, 755, 756, 758, 760, 762, 763, 766, 767, 768, 769, 770, 772, 776, 778, 784, 785, 787, 788, 789, 791, 792, 793, 794, 795, 797, 799, 803, 804, 805, 806, 814, 816, 820, 821, 822, 825, 828, 829, 830, 834, 837, 838, 842, 846, 851, 853, 863, 864, 866, 867, 869, 871, 878, 879, 880, 881, 882, 884, 885, 886, 888, 889, 891, 892, 893, 894, 895, 896, 898, 900, 901, 902, 906, 909, 912, 915, 916, 917, 918, 920, 921, 922, 923, 926, 928, 929, 931, 932, 933, 934, 935, 937, 940]
|
notebooks/create_models.ipynb | ###Markdown
Create model classes
###Code
# Make conv layer class for easy writing in next class
class ConvLayer(nn.Module):
def __init__(self, in_channels, out_channels, kernel_size, stride=1):
super().__init__()
# We have to keep the image size same
num_pad = int(np.floor(kernel_size / 2))
self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, stride, padding=num_pad)
def forward(self, x):
return self.conv(x)
class BottleneckBlock(nn.Module):
"""
Bottleneck layer similar to resnet bottleneck layer. InstanceNorm is used
instead of BatchNorm because when we want to generate images, we normalize
all the images independently.
(In batch norm you compute mean and std over complete batch, while in instance
norm you compute mean and std of each image independently). The reason for
doing this is, the generated images are independent of each other, so we should
not normalize them using a common statistic.
If you confused about the bottleneck architecture refer to the official pytorch
resnet implementation and paper.
"""
def __init__(self, in_channels, out_channels, kernel_size=3, stride=1):
super().__init__()
self.in_c = in_channels
self.out_c = out_channels
self.identity_block = nn.Sequential(
ConvLayer(in_channels, out_channels//4, kernel_size=1, stride=1),
nn.InstanceNorm2d(out_channels//4),
nn.ReLU(),
ConvLayer(out_channels//4, out_channels//4, kernel_size, stride=stride),
nn.InstanceNorm2d(out_channels//4),
nn.ReLU(),
ConvLayer(out_channels//4, out_channels, kernel_size=1, stride=1),
nn.InstanceNorm2d(out_channels),
nn.ReLU(),
)
self.shortcut = nn.Sequential(
ConvLayer(in_channels, out_channels, 1, stride),
nn.InstanceNorm2d(out_channels),
)
def forward(self, x):
out = self.identity_block(x)
if self.in_c == self.out_c:
residual = x
else:
residual = self.shortcut(x)
out += residual
out = F.relu(out)
return out
# Not used in the implementation
class UpSample(nn.Module):
def __init__(self, in_channels, out_channels, scale_factor, mode='bilinear'):
super().__init__()
self.scale_factor = scale_factor
self.mode = mode
self.conv = nn.Conv2d(in_channels, out_channels, kernel_size=3, stride=2, padding=1)
self.norm = nn.InstanceNorm2d(out_channels)
def forward(self, x):
out = self.conv(x)
out = F.interpolate(out, scale_factor=self.scale_factor, mode=self.mode, align_corners=False)
out = self.norm(out)
out = F.relu(out)
return out
# Helper functions for HRNet
def conv_down(in_c, out_c, stride=2):
return nn.Conv2d(in_c, out_c, kernel_size=3, stride=stride, padding=1)
def upsample(scale_factor):
return nn.Upsample(scale_factor=scale_factor, mode='bilinear')
class HRNet(nn.Module):
"""
For model reference see Figure 2 of the paper https://arxiv.org/pdf/1904.11617v1.pdf.
Naming convention used.
I refer to vertical layers as a single layer, so from left to right we have 8 layers
excluding the input image.
E.g. layer 1 contains the 500x500x16 block
layer 2 contains 500x500x32 and 250x250x32 blocks and so on
self.layer{x}_{y}:
x :- the layer number, as explained above
y :- the index number for that function starting from 1. So if layer 3 has two
downsample functions I write them as `downsample3_1`, `downsample3_2`
"""
def __init__(self):
super().__init__()
self.layer1_1 = BottleneckBlock(3, 16)
self.layer2_1 = BottleneckBlock(16, 32)
self.downsample2_1 = conv_down(16, 32)
self.layer3_1 = BottleneckBlock(32, 32)
self.layer3_2 = BottleneckBlock(32, 32)
self.downsample3_1 = conv_down(32, 32)
self.downsample3_2 = conv_down(32, 32, stride=4)
self.downsample3_3 = conv_down(32, 32)
self.layer4_1 = BottleneckBlock(64, 64)
self.layer5_1 = BottleneckBlock(192, 64)
self.layer6_1 = BottleneckBlock(64, 32)
self.layer7_1 = BottleneckBlock(32, 16)
self.layer8_1 = conv_down(16, 3, stride=1)
def forward(self, x):
map1_1 = self.layer1_1(x)
map2_1 = self.layer2_1(map1_1)
map2_2 = self.downsample2_1(map1_1)
map3_1 = torch.cat((self.layer3_1(map2_1), upsample(map2_2, 2)), 1)
map3_2 = torch.cat((self.downsample3_1(map2_1), self.layer3_2(map2_2)), 1)
map3_3 = torch.cat((self.downsample3_2(map2_1), self.downsample3_3(map2_2)), 1)
map4_1 = torch.cat((self.layer4_1(map3_1), upsample(map3_2, 2), upsample(map3_3, 4)), 1)
out = self.layer5_1(map4_1)
out = self.layer6_1(out)
out = self.layer7_1(out)
out = self.layer8_1(out)
return out
###Output
_____no_output_____
###Markdown
Create utility functions for image loading
###Code
def load_image(path, size=None):
"""
Resize img to size, size should be int and also normalize the
image using imagenet_stats
"""
img = Image.open(path)
if size is not None:
img = img.resize((size, size))
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))
])
img = transform(img).unsqueeze(0)
return img
def im_convert(img):
"""
Convert img from pytorch tensor to numpy array, so we can plot it.
It follows the standard method of denormalizing the img and clipping
the outputs
Input:
img :- (batch, channel, height, width)
Output:
img :- (height, width, channel)
"""
img = img.to('cpu').clone().detach()
img = img.numpy().squeeze(0)
img = img.transpose(1, 2, 0)
img = img * np.array((0.229, 0.224, 0.225)) + np.array((0.485, 0.456, 0.406))
img = img.clip(0, 1)
return img
def get_features(img, model, layers=None):
"""
Use VGG19 to extract features from the intermediate layers.
"""
if layers is None:
layers = {
'0': 'conv1_1', # style layer
'5': 'conv2_1', # style layer
'10': 'conv3_1', # style layer
'19': 'conv4_1', # style layer
'28': 'conv5_1', # style layer
'21': 'conv4_2' # content layer
}
features = {}
x = img
for name, layer in model._modules.items():
x = layer(x)
if name in layers:
features[layers[name]] = x
return features
def get_gram_matrix(img):
"""
Compute the gram matrix by converting to 2D tensor and doing dot product
img: (batch, channel, height, width)
"""
b, c, h, w = img.size()
img = img.view(b*c, h*w)
gram = torch.mm(img, img.t())
return gram
###Output
_____no_output_____
###Markdown
Write style_transfer.py Refer to train_model.ipynb for continuation of this notebook
###Code
# For data, place your images in the img folder and name it as content.png and style.png
# You can also input your images directly (for .py script)
class Args:
def __init__(self):
self.img_root = 'src/imgs'
self.content_img = 'content.png'
self.style_img = 'style.png'
self.use_batch = False
self.bs = 16
self.use_gpu = True
args = Args()
if args.use_gpu:
if torch.cuda.is_available():
device = torch.device('cuda')
else:
raise Exception('GPU is not available')
else:
device = torch.device('cpu')
# Load VGG19 features
vgg = vgg19(pretrained=True).features
vgg = vgg.to(device)
# We don't want to train VGG
for param in vgg.parameters():
param.requires_grad_(False)
# Load style net
style_net = HRNet()
style_net = style_net.to(device)
torch.backends.cudnn.benchmark = True
import os
content_img = load_image(os.path.join(args.img_root, args.content_img), size=500)
content_img = content_img.to(device)
style_img = load_image(os.path.join(args.img_root, args.style_img))
style_img = style_img.to(device)
content_img.size(), style_img.size()
###Output
_____no_output_____ |
test/math-test/notebook.ipynb | ###Markdown
This is an example project where I want to extract parameters from piece of spectrum data that I have. My gaussian function has the following form:$f(x) = a \mathrm{e}{\frac{-(x-c)^2}{2c^2}} + d$Where $a$ is a normalisation coefficient, $b$ is the center point, $c$ defines with the width of the curve and $d$ is the height above the x axis.First lets load and plot the data
###Code
%pylab inline
%config InlineBackend.figure_format = 'retina'
import numpy as np
x, y = np.loadtxt('./data/data.txt').T
plt.plot(x, y)
plt.xlabel('Wavelength [nm]')
plt.ylabel('Counts [a.u.]')
###Output
_____no_output_____
###Markdown
Clearly the data is a nice gaussian, so lets fit the function to get the center point and full-width at half max, which is given by $FWHM = 2\sqrt{2 ln(2)}$First we import some code which contains the gauss equation, and the optimize function from scipy to do the curve fitting.
###Code
from scipy.optimize import curve_fit
def gauss(t, a, b, c, d):
return a*np.exp(-((t-b)**2)/(2*c**2)) + d
p0 = [700, 990, 2, 0]
params, cov = curve_fit(gauss, x, y, p0=p0)
err = np.sqrt(np.diag(cov))
print("a = %lf +- %lf" % (params[0], err[0]))
print("b = %lf +- %lf" % (params[1], err[1]))
print("c = %lf +- %lf" % (params[2], err[2]))
print("d = %lf +- %lf" % (params[3], err[3]))
###Output
a = 9940.234967 +- 112.462988
b = 995.014369 +- 0.013048
c = 1.002707 +- 0.013202
d = 761.728754 +- 14.103390
|
matrix_one/matrix_day_4.ipynb | ###Markdown
Imports
###Code
import pandas as pd
import numpy as np
from sklearn.tree import DecisionTreeRegressor
from sklearn.metrics import mean_absolute_error
from sklearn.model_selection import cross_val_score
cd "/content/drive/My Drive/Colab Notebooks/dw_matrix/matrix_one"
df = pd.read_csv('data/shoes_prices.csv', low_memory=False)
df.shape
df.columns
mean_price = np.mean(df.prices_amountmin)
mean_price
[3] *5
y_true = df.prices_amountmin
y_pred = [mean_price] * y_true.shape[0]
mean_absolute_error(y_true, y_pred)
df.prices_amountmin.hist(bins=100)
np.log1p(df.prices_amountmin).hist(bins=100)
y_true = df.prices_amountmin
y_pred = [np.median(y_true)] * y_true.shape[0]
mean_absolute_error(y_true, y_pred)
y_true = df.prices_amountmin
price_log_mean = np.expm1(np.mean(np.log1p(y_true)))
y_pred = [price_log_mean] * y_true.shape[0]
mean_absolute_error(y_true, y_pred)
df.columns
df.brand.value_counts()
df['brand_cat'] = df.brand.factorize()[0]
df['manufacturer_cat'] = df.manufacturer.factorize()[0]
features = ['brand_cat']
def run_model(features):
X = df[features].values
y = df.prices_amountmin.values
model = DecisionTreeRegressor(max_depth=5)
scores = cross_val_score(model, X,y,scoring='neg_mean_absolute_error')
return np.mean(scores), np.std(scores)
run_model(['brand_cat'])
run_model(['manufacturer_cat'])
run_model(['brand_cat', 'manufacturer_cat'])
###Output
_____no_output_____ |
Learning_Basics.ipynb | ###Markdown
Tokenization
###Code
from nltk.tokenize import sent_tokenize, word_tokenize
nltk.download('punkt')
document = """
We are having fun at major league hacking local hack day build!
feel free to join the discord channel. also share about the session on social media.
"""
sentence = "send all the documents related to chapter 1,2,3,4 to [email protected]"
sents = sent_tokenize(document)
words = word_tokenize?
words = word_tokenize
###Output
_____no_output_____
###Markdown
words Stopward removal
###Code
from nltk.corpus import stopwords
nltk.download('stopwords')
sw = set(stopwords.words("english"))
print("has" in sw)
sentence = "send all the documents related to chapter 1,2,3,4 to [email protected]".split()
print(sentence)
def remove_stopwords(text, stopwords):
useful_words = [word for word in text if word not in stopwords]
return useful_words
remove_stopwords(sentence, sw)
sentence = "send all the documents related to chapter 1,2,3,4 to [email protected]"
from nltk.tokenize import RegexpTokenizer
tokenizer = RegexpTokenizer('[a-zA-Z@.]+')
useful_text = tokenizer.tokenize(sentence)
print(useful_text)
###Output
['send', 'all', 'the', 'documents', 'related', 'to', 'chapter', 'to', '[email protected]']
###Markdown
Stemming
###Code
document = """
I like this session and I liked all the sessions. I am liking everything.
We are having fun at major league hacking local hack day build!
feel free to join the discord channel. also share about the session on social media.
"""
from nltk.stem import PorterStemmer
ps = PorterStemmer()
ps.stem('liking')
corpus = [
'Indian cricket team will win World Cup, says Capt. Virat Kohli. World cup will be held at Sri Lanka.',
'We will win next Lok Sabha Elections, says confident Indian PM',
'The nobel laurate won the hearts of the people',
'The movie Raazi is an exciting Indian Spy thriller based upon a real story'
]
from sklearn.feature_extraction.text import CountVectorizer
cv = CountVectorizer()
vectorized_corpus = cv.fit_transform(corpus)
vectorized_corpus = vectorized_corpus.toarray()
print(vectorized_corpus[3])
print(cv.vocabulary_)
###Output
{'indian': 12, 'cricket': 6, 'team': 31, 'will': 37, 'win': 38, 'world': 40, 'cup': 7, 'says': 27, 'capt': 4, 'virat': 35, 'kohli': 14, 'be': 3, 'held': 11, 'at': 1, 'sri': 29, 'lanka': 15, 'we': 36, 'next': 19, 'lok': 17, 'sabha': 26, 'elections': 8, 'confident': 5, 'pm': 23, 'the': 32, 'nobel': 20, 'laurate': 16, 'won': 39, 'hearts': 10, 'of': 21, 'people': 22, 'movie': 18, 'raazi': 24, 'is': 13, 'an': 0, 'exciting': 9, 'spy': 28, 'thriller': 33, 'based': 2, 'upon': 34, 'real': 25, 'story': 30}
###Markdown
vectorisation with stopword removal
###Code
def myTokenizer(document):
words = tokenizer.tokenize(document.lower()) # this is nltk one
# remove stopwards
words = remove_stopwords(words, sw)
return words
myTokenizer('this is some function')
cv = CountVectorizer(tokenizer=myTokenizer)
vectorizedCorpus = cv.fit_transform(corpus).toarray()
print(len(vectorizedCorpus[0]))
print(cv.vocabulary_)
###Output
_____no_output_____ |
ml_course/ipynbfiles/MCQ1.ipynb | ###Markdown
Microsoft Python Exam Preparation MCQ Set1: 20 Questions--- **Q1 What Will Be The Output Of The Following Code Snippet?**```pythona=[1,2,3,4,5,6,7,8,9]print(a[::2])```1. [1,2] 2. [8,9]3. [1,3,5,7,9]4. [1,2,3] >Answer: *3. [1,3,5,7,9]*
###Code
a=[1,2,3,4,5,6,7,8,9]
print(a[::2])
###Output
[1, 3, 5, 7, 9]
###Markdown
---**Q2 Which of the following statements create a dictionary?**1. d = {}2. d = {'john':40, 'peter':45}3. d = {40:'john', 45:'peter'}4. All of the mentioned>Answer: *4. All of the mentioned*
###Code
d1 = {}
d2 = {"john":40, "peter":45}
d3 = {40:"john", 45:"peter"}
print(type(d1),type(d2),type(d3))
###Output
<class 'dict'> <class 'dict'> <class 'dict'>
###Markdown
---**Q3 Read the code shown below carefully and pick out the keys?**```pythond = {"john":40, "peter":45}```1. “john”, 40, 45, and “peter”2. “john” and “peter”3. 40 and 454. d = (40:”john”, 45:”peter”)>Answer: *2. “john” and “peter”*
###Code
d = {"john":40, "peter":45}
d.keys()
###Output
_____no_output_____
###Markdown
---**Q4 What Will Be The Output Of The Following Code Snippet?**```pythona=[1,2,3,4,5]print(a[3:0:-1])```1. Syntax error2. [4, 3, 2]3. [4, 3]4. [4, 3, 2, 1]>Answer: *2. [4,3,2]*
###Code
a=[1,2,3,4,5]
print(a[3:0:-1])
###Output
[4, 3, 2]
###Markdown
---**Q5 What Will Be TheOutput Of The Following Code Snippet?**```python init_tuple_a = 'a', 'b'init_tuple_b = ('a', 'b')print (init_tuple_a == init_tuple_b)```1. 02. 13. False4. True>Answer: *4. True*
###Code
init_tuple_a = 'a', 'b'
init_tuple_b = ('a', 'b')
print (init_tuple_a == init_tuple_b)
###Output
True
###Markdown
---**Q6 What Will Be The Output Of The Following Code Snippet?**```pythoninit_tuple_a = '1', '2'init_tuple_b = ('3', '4')print (init_tuple_a + init_tuple_b)```1. (1, 2, 3, 4)2. (‘1’, ‘2’, ‘3’, ‘4’)3. [‘1’, ‘2’, ‘3’, ‘4’]4. None>Answer: *2. ('1','2','3','4')*
###Code
init_tuple_a = '1', '2'
init_tuple_b = ('3', '4')
print (init_tuple_a + init_tuple_b)
###Output
('1', '2', '3', '4')
###Markdown
---**Q7 What Will Be The Output Of The Following Code Snippet?**```pythoninit_tuple_a = 1, 2init_tuple_b = (3, 4)[print(sum(x)) for x in [init_tuple_a + init_tuple_b]]```1. Nothing gets printed.2. 43. 104. TypeError: unsupported operand type>Answer: *3. 10*
###Code
init_tuple_a = 1, 2
init_tuple_b = (3, 4)
[print(sum(x)) for x in [init_tuple_a + init_tuple_b]]
###Output
10
###Markdown
---**Q8 What will be the output?**```pythond = {"john":40, "peter":45}print(list(d.keys()))```1. [“john”, “peter”]2. [“john”:40, “peter”:45]3. (“john”, “peter”)4. (“john”:40, “peter”:45)>Answer:*1. [“john”, “peter”]*
###Code
d = {"john":40, "peter":45}
print(list(d.keys()))
###Output
['john', 'peter']
###Markdown
---**Q9 You have a list which contains ten elements. Which of the following uses of range() would produce a list of the indexes in the list?**1. range(1,10)2. range(10)3. range(0,9)4. range(1,9)>Answer: *2. range(10)*
###Code
for i in range(10):
print(i,end=';')
###Output
0;1;2;3;4;5;6;7;8;9;
###Markdown
---**Q10 What is the output of the following?**```python i = 2while True: if i%3 == 0: break print(i) i += 2```1. 2 4 6 8 10 2. 2 4 3. 2 34. error>Answer: *2. 2 4*
###Code
i = 2
while True:
if i%3 == 0:
break
print(i)
i += 2
###Output
2
4
###Markdown
---**Q11 What is the output of the following?**```pythonx = "abcdef"i = "i"while i in x: print(i, end=" ")``` 1. no output 2. i i i i i i ... 3. a b c d e f 4. abcdef>Answer: *1. no output*
###Code
x = "abcdef"
i = "i"
while i in x:
print(i, end=" ")
###Output
_____no_output_____
###Markdown
---**Q12 What is the output of the following?**```pythonx = 'abcd'for i in x: print(i.upper())```1. a b c d 2. A B C D 3. a B C D 4. error>Answer: *2. A B C D*
###Code
x = 'abcd'
for i in x:
print(i.upper())
###Output
A
B
C
D
###Markdown
---**Q13 What is the output of the following code?**```pythonnums = set([1,1,2,3,3,3,4,4])print(len(nums))```1. 72. Error, invalid syntax for formation of set3. 44. 8>Answer: *3. 4*
###Code
nums = set([1,1,2,3,3,3,4,4])
print(len(nums))
###Output
4
###Markdown
---**Q14 Which of the following statements is used to create an empty set?**1. { }2. set()3. [ ]4. ( )>Answer: *2. set()*
###Code
x = {}
y = []
z = ()
m = set()
print(type(x),type(y),type(z),type(m))
###Output
<class 'dict'> <class 'list'> <class 'tuple'> <class 'set'>
###Markdown
---**Q15 Suppose t = (1, 2, 4, 3), which of the following is incorrect?**1. print(t[3])2. t[3] = 453. print(max(t))4. print(len(t))>Answer: *2. t[3] = 45*
###Code
t = (1, 2, 4, 3)
t[3]=9
###Output
_____no_output_____
###Markdown
---**Q16 What does the following code print to the console?**```pythonif False: print("Nissan")elif True: print("Ford")elif True: print("BMW")else: print("Audi")``` 1. Ford2. Ford BMW3. BMW4. None Of The Above>Answer: *1. Ford*
###Code
if False:
print("Nissan")
elif True:
print("Ford")
elif True:
print("BMW")
else:
print("Audi")
###Output
Ford
###Markdown
---**Q17 Suppose list1 is [1, 3, 2], What is list1 * 2 ?**1. [2, 6, 4] 2. [1, 3, 2, 1, 3] 3. [1, 3, 2, 1, 3, 2] 4. [1, 3, 2, 3, 2, 1]>Answer: *3. [1, 3, 2, 1, 3, 2]*
###Code
list1 = [1,3,2]
list1*2
###Output
_____no_output_____
###Markdown
---**Q18 What is the output when we execute list(“hello”)?**1. [‘h’, ‘e’, ‘l’, ‘l’, ‘o’]2. [‘hello’]3. [‘llo’]4. [‘olleh’]>Answer: *1. [‘h’, ‘e’, ‘l’, ‘l’, ‘o’]*
###Code
list('hello')
###Output
_____no_output_____
###Markdown
---**Q19 Suppose list1 is [2, 33, 222, 14, 25], What is list1[:-1] ?**1. [2, 33, 222, 14]2. Error3. 254. [25, 14, 222, 33, 2]>Answer: *1. [2, 33, 222, 14]*
###Code
list1 = [2, 33, 222, 14, 25]
list1[:-1]
###Output
_____no_output_____
###Markdown
---**Q-20 Suppose list1 is [4, 2, 2, 4, 5, 2, 1, 0], Which of the following is correct syntax for slicing operation ?**1. print(list1[0])2. print(list1[:2])3. print(list1[:-2])4. all of the mentioned>Answer: *4. all of the mentioned*
###Code
list1 = [4, 2, 2, 4, 5, 2, 1, 0]
print(list1[0])
print(list1[:2])
print(list1[:-2])
###Output
4
[4, 2]
[4, 2, 2, 4, 5, 2]
|
module2/LS_DS9_122_Sampling_Confidence_Intervals_and_Hypothesis_Testing_Assignment.ipynb | ###Markdown
Assignment - Build a confidence intervalA confidence interval refers to a neighborhood around some point estimate, the size of which is determined by the desired p-value. For instance, we might say that 52% of Americans prefer tacos to burritos, with a 95% confidence interval of +/- 5%.52% (0.52) is the point estimate, and +/- 5% (the interval $[0.47, 0.57]$) is the confidence interval. "95% confidence" means a p-value $\leq 1 - 0.95 = 0.05$.In this case, the confidence interval includes $0.5$ - which is the natural null hypothesis (that half of Americans prefer tacos and half burritos, thus there is no clear favorite). So in this case, we could use the confidence interval to report that we've failed to reject the null hypothesis.But providing the full analysis with a confidence interval, including a graphical representation of it, can be a helpful and powerful way to tell your story. Done well, it is also more intuitive to a layperson than simply saying "fail to reject the null hypothesis" - it shows that in fact the data does *not* give a single clear result (the point estimate) but a whole range of possibilities.maiormarso.comHow is a confidence interval built, and how should it be interpreted? It does *not* mean that 95% of the data lies in that interval - instead, the frequentist interpretation is "if we were to repeat this experiment 100 times, we would expect the average result to lie in this interval ~95 times."For a 95% confidence interval and a normal(-ish) distribution, you can simply remember that +/-2 standard deviations contains 95% of the probability mass, and so the 95% confidence interval based on a given sample is centered at the mean (point estimate) and has a range of +/- 2 (or technically 1.96) standard deviations.Different distributions/assumptions (90% confidence, 99% confidence) will require different math, but the overall process and interpretation (with a frequentist approach) will be the same.Your assignment - using the data from the prior module ([congressional voting records](https://archive.ics.uci.edu/ml/datasets/Congressional+Voting+Records)): Confidence Intervals:1. Generate and numerically represent a confidence interval2. Graphically (with a plot) represent the confidence interval3. Interpret the confidence interval - what does it tell you about the data and its distribution? Chi-squared tests:4. Take a dataset that we have used in the past in class that has **categorical** variables. Pick two of those categorical variables and run a chi-squared tests on that data - By hand using Numpy - In a single line using Scipy
###Code
!wget https://archive.ics.uci.edu/ml/machine-learning-databases/voting-records/house-votes-84.data
import pandas as pd
cols = [
'party',
'handicapped-infants',
'water-project',
'budget',
'physician-fee-freeze',
'el-salvador-aid',
'religious-groups',
'anti-satellite-ban',
'aid-to-contras',
'mx-missile',
'immigration',
'synfuels',
'education',
'right-to-sue',
'crime',
'duty_free',
'south_Africa']
df = pd.read_csv('house-votes-84.data', names=cols)
df.head(1)
import numpy as np
import pandas as pd
df=df.replace({'?': 'NaN', 'n':0.0, 'y':1.0,'republican':1,'democrat':0})
#df=df.replace({'?':np.NaN}maior)
df.head(8)
df.index
df.shape
df.party.value_counts()
df = df.astype(float)
rep = df[df['party'] ==1]
rep.head()
rep = rep.astype(float)
# rep.sum(axis = 0, skipna = True)
dem = df[df['party'] ==0]
dem.head()
dem = dem.astype(float)
# dem.sum(axis = 0, skipna = True)
df = df.astype(float)
from scipy.stats import t
from scipy import stats
CI = t.interval(0.95,df['budget'])
a = df['budget'].dropna()
confidence_interval = t.interval(0.95, len(a)-1, loc=np.mean(a), scale=stats.sem(a))
confidence_interval
budget_m = df['budget'].mean()
print(budget_m)
mean_dem_budget = dem['budget'].mean()
print(mean_dem_budget)
std_error_dem_budget = stats.sem(dem['budget'], nan_policy='omit')
print(std_error_dem_budget)
t_stat_dem_budget = stats.ttest_1samp(dem['budget'], .5, nan_policy='omit')
print(t_stat_dem_budget)
t_stat_dem_budget[0]
CI_plus = mean_dem_budget + t_stat_dem_budget[0]*std_error_dem_budget
print(CI_plus)
CI_minus = mean_dem_budget - t_stat_dem_budget[0]*std_error_dem_budget
print(CI_minus)
import seaborn as sns
import matplotlib.pyplot as plt
sns.kdeplot(dem['budget'])
plt.axvline(x=CI_plus, color='red')
plt.axvline(x=CI_minus, color='red')
plt.show()
###Output
/usr/local/lib/python3.6/dist-packages/statsmodels/nonparametric/kde.py:447: RuntimeWarning: invalid value encountered in greater
X = X[np.logical_and(X > clip[0], X < clip[1])] # won't work for two columns.
/usr/local/lib/python3.6/dist-packages/statsmodels/nonparametric/kde.py:447: RuntimeWarning: invalid value encountered in less
X = X[np.logical_and(X > clip[0], X < clip[1])] # won't work for two columns.
###Markdown
3.1 what does it tell you about the data and its distribution?1. My null Hypothosis is that they are voting evenly.2. Alternative would also be voting even.3. The confidence level is .95 4. Using the mean for the sample mean, there is no bell and the mean line is standing vulnerable with the cutoff lines in an awkward position.See 3.2 below
###Code
confidence = 0.95
(1 + confidence) / 2.0 # This converts confidence to two-tailed
confidence_level = .95
dof = 431 - 1
stats.t.ppf((1 + confidence_level) / 2, dof)
original_sample=a
sample_means = []
for x in range(3000):
m = np.random.choice(original_sample,300).mean()
sample_means.append(m)
import seaborn as sns
import matplotlib.pyplot as plt
sns.distplot(sample_means)
plt.axvline(x=original_sample.mean())
plt.axvline(x=confidence_interval[0], color='r')
plt.axvline(x=confidence_interval[1], color='r')
plt.xlim(0, 1)
df['budget'].value_counts()
###Output
_____no_output_____
###Markdown
3.2 what does it tell you about the data and its distribution?1. My null Hypothosis is that they are voting evenly.2. Alternative would also be voting even.3. The confidence level is .95 4. Using 3000 for the sample mean, the bell stands straight up narrow towering tightly between the cutoff lines.
###Code
#m[0][1][1]
sample_means = []
for x in range(3000):
m = np.random.choice(original_sample,300).mean()
sample_means.append(m)
from scipy.stats import bayes_mvs
m=bayes_mvs(original_sample)
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline
import matplotlib.pyplot as plt
sns.distplot(sample_means)
plt.axvline(x=m[0][0], color='r')
plt.axvline(x=m[0][1][0], color='r')
plt.axvline(x=m[0][1][1], color='r')
plt.title('Budget Bill')
plt.ylabel('votes')
plt.xlim(0, 1)
#dem['budget'].plot.hist();
df = df.astype(float)
demlist = dem["budget"].tolist()
demlist
print(*demlist, sep=",")
replist = rep["budget"].tolist()
replist
print(*replist, sep = ",")
ddf=df[['party','budget','crime']]
ddf
import pandas as pd
dc=df[['party','budget','crime']]
ddf[['budget','crime']]
contingency_table = pd.crosstab(df['party'], df['budget'])
contingency_table
e1 = (contingency_table[0.0].sum()/contingency_table.sum().sum())*contingency_table.loc[0.0].sum()
e2 = (contingency_table[1.0].sum()/contingency_table.sum().sum())*contingency_table.loc[0.0].sum()
e3 = (contingency_table[0.0].sum()/contingency_table.sum().sum())*contingency_table.loc[1.0].sum()
e4 = (contingency_table[1.0].sum()/contingency_table.sum().sum())*contingency_table.loc[1.0].sum()
contingency_table = pd.crosstab(df['party'], df['budget'])
contingency_table
((contingency_table[0.0][0.0]-(contingency_table[0.0].sum()/contingency_table.sum().sum())*contingency_table.loc[0.0].sum() )**2) / ((contingency_table[0.0].sum()/contingency_table.sum().sum())*contingency_table.loc[0.0].sum())
((contingency_table[1.0][0.0]-(contingency_table[1.0].sum()/contingency_table.sum().sum())*contingency_table.loc[0.0].sum() )**2) / ((contingency_table[1.0].sum()/contingency_table.sum().sum())*contingency_table.loc[0.0].sum())
((contingency_table[0.0][1.0]-(contingency_table[0.0].sum()/contingency_table.sum().sum())*contingency_table.loc[1.0].sum() )**2) / ((contingency_table[0.0].sum()/contingency_table.sum().sum())*contingency_table.loc[1.0].sum())
((contingency_table[1.0][1.0]-(contingency_table[1.0].sum()/contingency_table.sum().sum())*contingency_table.loc[1.0].sum() )**2) / ((contingency_table[1.0].sum()/contingency_table.sum().sum())*contingency_table.loc[1.0].sum()) #maior
54.878823449528504+37.09201110620308+87.00301278583787+58.80440785129753
chi_squared, p_value, dof, expected = stats.chi2_contingency(contingency_table)
print(f"Chi-Squared: {chi_squared}")
print(f"P-value: {p_value}")
print(f"Degrees of Freedom: {dof}")
print("Expected: \n", np.array(expected))
###Output
Chi-Squared: 234.65408769323486
P-value: 5.759792112623893e-53
Degrees of Freedom: 1
Expected:
[[104.85849057 155.14150943]
[ 66.14150943 97.85849057]]
|
Kaggle_Challenge_X7.ipynb | ###Markdown
###Code
# Installs
%%capture
!pip install --upgrade category_encoders plotly
# Imports
import os, sys
os.chdir('/content')
!git init .
!git remote add origin https://github.com/LambdaSchool/DS-Unit-2-Kaggle-Challenge.git
!git pull origin master
!pip install -r requirements.txt
os.chdir('module1')
# Imports
import pandas as pd
import numpy as np
import math
import sklearn
sklearn.__version__
from sklearn.model_selection import train_test_split
# Import the models
from sklearn.linear_model import LogisticRegressionCV
from sklearn.pipeline import make_pipeline
# Import encoder and scaler and imputer
import category_encoders as ce
from sklearn.preprocessing import StandardScaler
from sklearn.impute import SimpleImputer
# Import random forest classifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score
def main_program():
def wrangle(X):
# Wrangles train, validate, and test sets
X = X.copy()
# Convert date_recorded to datetime
X['date_recorded'] = pd.to_datetime(X['date_recorded'], infer_datetime_format=True)
# Extract components from date_recorded and drop the original column
X['year_recorded'] = X['date_recorded'].dt.year
X['month_recorded'] = X['date_recorded'].dt.month
X['day_recorded'] = X['date_recorded'].dt.day
X = X.drop(columns='date_recorded')
# Engineer new feature years - construction_year to date_recorded
X.loc[X['construction_year'] == 0, 'construction_year'] = np.nan
X['years'] = X['year_recorded'] - X['construction_year']
# Remove latitude outliers
X['latitude'] = X['latitude'].replace(-2e-08, np.nan)
# Features with many zero's are likely nan's
cols_with_zeros = ['construction_year', 'longitude', 'latitude', 'gps_height',
'population', 'amount_tsh']
for col in cols_with_zeros:
X[col] = X[col].replace(0, np.nan)
# Impute mean for years
X.loc[X['years'].isna(), 'years'] = X['years'].mean()
#X.loc[X['pump_age'].isna(), 'pump_age'] = X['pump_age'].mean()
# Impute mean for longitude and latitude based on region
average_lat = X.groupby('region').latitude.mean().reset_index()
average_long = X.groupby('region').longitude.mean().reset_index()
shinyanga_lat = average_lat.loc[average_lat['region'] == 'Shinyanga', 'latitude']
shinyanga_long = average_long.loc[average_long['region'] == 'Shinyanga', 'longitude']
X.loc[(X['region'] == 'Shinyanga') & (X['latitude'] > -1), ['latitude']] = shinyanga_lat[17]
X.loc[(X['region'] == 'Shinyanga') & (X['longitude'].isna()), ['longitude']] = shinyanga_long[17]
mwanza_lat = average_lat.loc[average_lat['region'] == 'Mwanza', 'latitude']
mwanza_long = average_long.loc[average_long['region'] == 'Mwanza', 'longitude']
X.loc[(X['region'] == 'Mwanza') & (X['latitude'] > -1), ['latitude']] = mwanza_lat[13]
X.loc[(X['region'] == 'Mwanza') & (X['longitude'].isna()) , ['longitude']] = mwanza_long[13]
#X.loc[X['amount_tsh'].isna(), 'amount_tsh'] = 0
# Clean installer
X['installer'] = X['installer'].str.lower()
X['installer'] = X['installer'].str[:4]
X['installer'].value_counts(normalize=True)
tops = X['installer'].value_counts()[:15].index
X.loc[~X['installer'].isin(tops), 'installer'] = 'other'
# Bin lga
#tops = X['lga'].value_counts()[:10].index
#X.loc[~X['lga'].isin(tops), 'lga'] = 'Other'
# Bin subvillage
tops = X['subvillage'].value_counts()[:25].index
X.loc[~X['subvillage'].isin(tops), 'subvillage'] = 'Other'
# Impute mean for a feature based on latitude and longitude
def latlong_conversion(feature, pop, long, lat):
radius = 0.1
radius_increment = 0.3
if math.isnan(pop):
pop_temp = 0
while pop_temp <= 1 and radius <= 2:
lat_from = lat - radius
lat_to = lat + radius
long_from = long - radius
long_to = long + radius
df = X[(X['latitude'] >= lat_from) &
(X['latitude'] <= lat_to) &
(X['longitude'] >= long_from) &
(X['longitude'] <= long_to)]
pop_temp = df[feature].mean()
radius = radius + radius_increment
else:
pop_temp = pop
if np.isnan(pop_temp):
new_pop = X_train[feature].mean()
else:
new_pop = pop_temp
return new_pop
X.loc[X['population'].isna(), 'population'] = X['population'].mean()
#X['population'] = X.apply(lambda x: latlong_conversion('population', x['population'], x['longitude'], x['latitude']), axis=1)
# Impute mean for tsh based on mean of source_class/basin/waterpoint_type_group
#def tsh_calc(tsh, source, base, waterpoint):
# if math.isnan(tsh):
# if (source, base, waterpoint) in tsh_dict:
# new_tsh = tsh_dict[source, base, waterpoint]
# return new_tsh
# else:
# return tsh
# return tsh
#temp = X[~X['amount_tsh'].isna()].groupby(['source_class',
# 'basin',
# 'waterpoint_type_group'])['amount_tsh'].mean()
#tsh_dict = dict(temp)
#X['amount_tsh'] = X.apply(lambda x: tsh_calc(x['amount_tsh'], x['source_class'], x['basin'], x['waterpoint_type_group']), axis=1)
# Drop unneeded columns
unusable_variance = ['recorded_by', 'id', 'num_private', 'wpt_name']
X = X.drop(columns=unusable_variance)
return X
# Merge train_features.csv & train_labels.csv
train = pd.merge(pd.read_csv('../data/tanzania/train_features.csv'),
pd.read_csv('../data/tanzania/train_labels.csv'))
# Read test_features.csv & sample_submission.csv
test = pd.read_csv('../data/tanzania/test_features.csv')
sample_submission = pd.read_csv('../data/tanzania/sample_submission.csv')
# Split train into train & val. Make val the same size as test.
target = 'status_group'
train, val = train_test_split(train, train_size=0.99, test_size=0.01,
stratify=train[target], random_state=42)
# Wrangle train, validate, and test sets in the same way
train = wrangle(train)
val = wrangle(val)
test = wrangle(test)
# Arrange data into X features matrix and y target vector
X_train = train.drop(columns=target)
y_train = train[target]
X_val = val.drop(columns=target)
y_val = val[target]
X_test = test
# Make pipeline!
pipeline = make_pipeline(
ce.OrdinalEncoder(),
SimpleImputer(strategy='mean'),
RandomForestClassifier(n_estimators=1400,
random_state=42,
min_samples_split=5,
min_samples_leaf=1,
max_features='auto',
max_depth=30,
bootstrap=True,
n_jobs=-1,
verbose = 1)
)
# Fit on train, score on val
pipeline.fit(X_train, y_train)
y_pred = pipeline.predict(X_val)
print('Validation Accuracy', accuracy_score(y_val, y_pred))
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import RandomizedSearchCV
# Number of trees in random forest
n_estimators = [int(x) for x in np.linspace(start = 200, stop = 2000, num = 10)]
# Number of features to consider at every split
max_features = ['auto', 'sqrt']
# Maximum number of levels in tree
max_depth = [int(x) for x in np.linspace(10, 110, num = 11)]
max_depth.append(None)
# Minimum number of samples required to split a node
min_samples_split = [2, 5, 10]
# Minimum number of samples required at each leaf node
min_samples_leaf = [1, 2, 4]
# Method of selecting samples for training each tree
bootstrap = [True, False]
# Create the random grid
random_grid = {'n_estimators': n_estimators,
'max_features': max_features,
'max_depth': max_depth,
'min_samples_split': min_samples_split,
'min_samples_leaf': min_samples_leaf,
'bootstrap': bootstrap}
print(random_grid)
y_train = np.where(y_train == 'functional', 1, y_train)
y_train = np.where(y_train == 'non functional', 2, y_train)
y_train = np.where(y_train == 'functional needs repair', 3, y_train)
pipeline = make_pipeline (
ce.OrdinalEncoder(),
SimpleImputer(strategy='mean'),
RandomizedSearchCV(estimator = RandomForestRegressor(),
param_distributions = random_grid,
n_iter = 5,
verbose=2,
random_state=42,
n_jobs = -1)
)
pipeline.fit(X_train, y_train)
pd.set_option('display.max_rows', 200)
model = pipeline.named_steps['randomizedsearchcv']
best = pd.Series(model.best_params_)
print(best)
# pd.set_option('display.max_rows', 200)
# model = pipeline.named_steps['randomforestclassifier']
# encoder = pipeline.named_steps['ordinalencoder']
# encoded_columns = encoder.transform(X_train).columns
# importances = pd.Series(model.feature_importances_, encoded_columns)
# importances.sort_values(ascending=False)
# assert all(X_test.columns == X_train.columns)
# y_pred = pipeline.predict(X_test)
# submission = sample_submission.copy()
# submission['status_group'] = y_pred
# submission.to_csv('/content/submission-f8.csv', index=False)
main_program()
#for i in range(3, 10):
# main_program(i)
#for i in range(1, 6):
# i = i * 5
# print('lga bins: ', i)
# main_program(i)
#i = 25
#for j in range(2, 6):
# j = j * 5
# for k in range(2, 6):
# k = k * 5
# print('installer bins: ', i, 'funder bins: ', j,'subvillage bins: ', k)
# main_program( i, j, k)
#pd.set_option('display.max_rows', 200)
#model = pipeline.named_steps['randomforestclassifier']
#encoder = pipeline.named_steps['ordinalencoder']
#encoded_columns = encoder.transform(X_train).columns
#importances = pd.Series(model.feature_importances_, encoded_columns)
#importances.sort_values(ascending=False)
#assert all(X_test.columns == X_train.columns)
#y_pred = pipeline.predict(X_test)
#submission = sample_submission.copy()
#submission['status_group'] = y_pred
#submission.to_csv('/content/submission-f2.csv', index=False)
###Output
_____no_output_____ |
Project4/Notebooks/AI2_HW4_parts_1_2_andreas_spanopoulos.ipynb | ###Markdown
Install useful libraries
###Code
!pip install sentence_transformers
###Output
Requirement already satisfied: sentence_transformers in /usr/local/lib/python3.6/dist-packages (0.4.1.2)
Requirement already satisfied: scipy in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (1.4.1)
Requirement already satisfied: tqdm in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (4.41.1)
Requirement already satisfied: transformers<5.0.0,>=3.1.0 in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (4.2.1)
Requirement already satisfied: scikit-learn in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (0.22.2.post1)
Requirement already satisfied: sentencepiece in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (0.1.95)
Requirement already satisfied: nltk in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (3.2.5)
Requirement already satisfied: torch>=1.6.0 in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (1.7.0+cu101)
Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from sentence_transformers) (1.19.5)
Requirement already satisfied: requests in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (2.23.0)
Requirement already satisfied: tokenizers==0.9.4 in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (0.9.4)
Requirement already satisfied: packaging in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (20.8)
Requirement already satisfied: regex!=2019.12.17 in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (2019.12.20)
Requirement already satisfied: dataclasses; python_version < "3.7" in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (0.8)
Requirement already satisfied: sacremoses in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (0.0.43)
Requirement already satisfied: importlib-metadata; python_version < "3.8" in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (3.3.0)
Requirement already satisfied: filelock in /usr/local/lib/python3.6/dist-packages (from transformers<5.0.0,>=3.1.0->sentence_transformers) (3.0.12)
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn->sentence_transformers) (1.0.0)
Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from nltk->sentence_transformers) (1.15.0)
Requirement already satisfied: typing-extensions in /usr/local/lib/python3.6/dist-packages (from torch>=1.6.0->sentence_transformers) (3.7.4.3)
Requirement already satisfied: future in /usr/local/lib/python3.6/dist-packages (from torch>=1.6.0->sentence_transformers) (0.16.0)
Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<5.0.0,>=3.1.0->sentence_transformers) (2020.12.5)
Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<5.0.0,>=3.1.0->sentence_transformers) (3.0.4)
Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<5.0.0,>=3.1.0->sentence_transformers) (1.24.3)
Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->transformers<5.0.0,>=3.1.0->sentence_transformers) (2.10)
Requirement already satisfied: pyparsing>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from packaging->transformers<5.0.0,>=3.1.0->sentence_transformers) (2.4.7)
Requirement already satisfied: click in /usr/local/lib/python3.6/dist-packages (from sacremoses->transformers<5.0.0,>=3.1.0->sentence_transformers) (7.1.2)
Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.6/dist-packages (from importlib-metadata; python_version < "3.8"->transformers<5.0.0,>=3.1.0->sentence_transformers) (3.4.0)
###Markdown
Imports
###Code
import os
import re
import time
import json
import torch
import pickle
import nltk
import nltk.data
nltk.download('punkt')
from sentence_transformers import SentenceTransformer, util
from preprocessing import *
###Output
_____no_output_____
###Markdown
Preprocessing Start by creating the class that will be used to store an article If the Dataset has already been parsed and stored in a pickle file, set this variable to True, to avoid parsing it again. If you are running this Notebook for the first time, then of course, the Dataset has not been parsed, so set the variable to False.
###Code
has_already_been_parsed = True
class CovidArticle:
""" class used to keep the information of an article """
def __init__(self, article_data):
tokenizer = nltk.data.load('tokenizers/punkt/english.pickle')
self._id = article_data["paper_id"]
self._name = article_data["metadata"]["title"]
self._abstract = self._preprocess_abstract(article_data["abstract"], tokenizer)
self._corpus_text = self._preprocess_bodytext(article_data["body_text"], tokenizer)
@property
def id(self):
return self._id
@property
def title(self):
return self._name
@property
def abstract(self):
return self._abstract
@property
def corpus(self):
return self._corpus_text
@id.setter
def id(self, _id):
self._id = _id
@title.setter
def title(self, _title):
self._name = _title
@abstract.setter
def abstract(self, _abstract):
self._abstract = _abstract
@corpus.setter
def corpus(self, _corpus):
self._corpus_text = _corpus
@staticmethod
def _preprocess(text_data):
result = remove_urls(text_data)
result = remove_references(result)
result = remove_multiple_full_stops(result)
result = remove_et_al(result)
result = remove_figure_references(result)
result = remove_multiple_whitespace(result)
return result
def _preprocess_abstract(self, abstract, tokenizer):
if not abstract:
sentences = [""]
else:
sentences = []
for paragraph in abstract:
sentences = sentences + self._preprocess_paragraph(paragraph, tokenizer)[1]
return ("Abstract.", sentences)
def _preprocess_bodytext(self, body_text, tokenizer):
return [self._preprocess_paragraph(paragraph, tokenizer)
for paragraph in body_text] if body_text else [("", "")]
def _preprocess_paragraph(self, paragraph, tokenizer):
return (paragraph['section'] + '.',
[sentence.strip()
for sentence in tokenizer.tokenize(self._preprocess(paragraph['text']))
if sentence != '.'])
@property
def summary(self):
""" returns a list containing the title + the sentences in the abtract plus section names """
sentences = [paragraph_and_section[0]
for paragraph_and_section in self._corpus_text
if paragraph_and_section[0] != '.']
return list(set([self.title + '.', self._abstract[0], *self._abstract[1]] + sentences))
@property
def text(self):
""" returns all the text in the article: abstract + sections + paragraph text """
sentences = [self._abstract[0], *self._abstract[1]]
for paragraph_and_section in self._corpus_text:
if paragraph_and_section[0] != '.':
sentences.append(paragraph_and_section[0])
if paragraph_and_section[1] != '.':
sentences = sentences + paragraph_and_section[1]
return sentences
def get_section_from_index(self, idx):
""" given an index of a sentence in the whole text of the article, it
returns the section and the index of the sentence inside the section """
text = self.text
abstract = self.abstract
corpus_text = self.corpus
length_of_abstract = 1 + len(abstract[1])
target_index = None
target_section = None
# indexed sentence is in abstract
if idx < length_of_abstract:
target_index = idx
target_section = abstract
# else, indexed sentence is in body text
else:
current_index = length_of_abstract
for paragraph in corpus_text:
paragraph_length = 1 + len(paragraph[1])
if idx <= current_index + paragraph_length:
target_index = idx - current_index
target_section = paragraph
break
else:
current_index += paragraph_length
return target_index, target_section
###Output
_____no_output_____
###Markdown
Helper method to parse all the articles
###Code
def get_articles(root_dir, filenames, log_every=None, stop_at=None):
"""
:param str root_dir: The root directory containing all the articles in json format.
:param list[str] filenames: A list containing the names of the json files.
:param int log_every: Frequency of prints that show how many files have been parsed at a
specific timestep.
:param int stop_at: The number of articles to parsed. If None, then all the articles
available will be parsed.
:return: A list containing Article objectes, one for every article.
:rtype: list[CovidArticle]
"""
covid_articles = []
for filename in filenames:
full_filepath = os.path.join(root_dir, filename)
with open(full_filepath) as f:
data = json.load(f)
covid_articles.append(CovidArticle(data))
if log_every is not None and len(covid_articles) % log_every == 0:
print('{} articles parsed'.format(len(covid_articles)))
if stop_at is not None and len(covid_articles) == stop_at:
break
return covid_articles
###Output
_____no_output_____
###Markdown
Get a list with the articles either by- Parsing it using the ``` get_articles() ``` function, which takes ~ 1 hour, and then saving it in a pickle file.- Loading an already parsed pickle file containing the list.I would suggest first parsing the Dataset in CPU, and then switching to GPU to load it.
###Code
# edit your paths here
root_dir = os.path.join('.', 'drive', 'My Drive', 'Colab Notebooks', 'AI2', 'Project4', 'Dataset', 'comm_use_subset')
save_path = os.path.join('.', 'drive', 'My Drive', 'Colab Notebooks', 'AI2', 'Project4', 'Preprocessed_Dataset', 'processed_articles.pickle')
# if the dataset is to be parsed for the first time
if not has_already_been_parsed:
filenames = sorted(os.listdir(root_dir))
articles = get_articles(root_dir, filenames, log_every=100)
with open(save_path, 'wb') as f:
pickle.dump(articles, f)
# else, it has already been parsed, just load it
else:
with open(save_path, 'rb') as f:
articles = pickle.load(f)
len(articles)
###Output
_____no_output_____
###Markdown
Let's take a look at IDs and the titles of the first 20 articles
###Code
for article in articles[:20]:
print('ID: {},\tTitle: {}'.format(article.id, article.title))
###Output
ID: 000b7d1517ceebb34e1e3e817695b6de03e2fa78, Title: Supplementary Information An eco-epidemiological study of Morbilli-related paramyxovirus infection in Madagascar bats reveals host-switching as the dominant macro-evolutionary mechanism
ID: 00142f93c18b07350be89e96372d240372437ed9, Title: immunity to pathogens taught by specialized human dendritic cell subsets
ID: 0022796bb2112abd2e6423ba2d57751db06049fb, Title: Public Health Responses to and Challenges for the Control of Dengue Transmission in High-Income Countries: Four Case Studies
ID: 00326efcca0852dc6e39dc6b7786267e1bc4f194, Title: a section of the journal Frontiers in Pediatrics A Review of Pediatric Critical Care in Resource-Limited Settings: A Look at Past, Present, and Future Directions
ID: 00352a58c8766861effed18a4b079d1683fec2ec, Title: MINI REVIEW Function of the Deubiquitinating Enzyme USP46 in the Nervous System and Its Regulation by WD40-Repeat Proteins
ID: 0043d044273b8eb1585d3a66061e9b4e03edc062, Title: Evaluation of the tuberculosis programme in Ningxia Hui Autonomous region, the People's Republic of China: a retrospective case study
ID: 0049ba8861864506e1e8559e7815f4de8b03dbed, Title: GPI-anchored single chain Fv -an effective way to capture transiently-exposed neutralization epitopes on HIV-1 envelope spike
ID: 00623bf2715e25d3acacb3f210d6888ed840e3cb, Title: Transmissible gastroenteritis virus infection decreases arginine uptake by downregulating CAT-1 expression
ID: 0072159e1ebecc889e9bcabb58bb45c47e18a403, Title: Chaperone-Mediated Autophagy Protein BAG3 Negatively Regulates Ebola and Marburg VP40-Mediated Egress
ID: 007618ad76a3548195ab5d11c1e2459931c91cd1, Title: Molecular Sciences Monocytes and Macrophages as Viral Targets and Reservoirs
ID: 007bf75961da42a7e0cc8e2855e5c208a5ec65c1, Title: The Murine Coronavirus Hemagglutinin-esterase Receptor-binding Site: A Major Shift in Ligand Specificity through Modest Changes in Architecture
ID: 0080d3bd9fb92e022c27715c2d1249042aa998b8, Title: Rational Design of a Live Attenuated Dengue Vaccine: 29-O-Methyltransferase Mutants Are Highly Attenuated and Immunogenic in Mice and Macaques
ID: 0089aa4b17549b9774f13a9e2e12a84fc827d60b, Title: The Domain-Specific and Temperature-Dependent Protein Misfolding Phenotype of Variant Medium-Chain acyl-CoA Dehydrogenase
ID: 008c1ceaeffe7abc87b031af39fae2632fa72897, Title: AMS 3.0: prediction of post-translational modifications
ID: 008d980cbcc283a9b707de3d9a02573dde8528ac, Title: A pilot study-genetic diversity and population structure of snow leopards of Gilgit-Baltistan, Pakistan, using molecular techniques
ID: 009002e8a66b8c1df088cf04069629fd76b13bb9, Title: Epidemiological Characteristics of Imported Influenza A (H1N1) Cases during the 2009 Pandemic in Korea
ID: 0093f9ae0861afc0d29fff935ae6a3af898cea00, Title: Emergence of infectious malignant thrombocytopenia in Japanese macaques (Macaca fuscata) by SRV-4 after transmission to a novel host
ID: 0094b25e2500306fadbdfb41d520f2970bb086d3, Title: BMC Infectious Diseases Sex-and age-dependent association of SLC11A1 polymorphisms with tuberculosis in Chinese: a case control study
ID: 00951716e01c8e0cc341770389fc38d1b5455210, Title: Knowledge of, attitudes toward, and preventive practices relating to cholera and oral cholera vaccine among urban high-risk groups: findings of a cross-sectional study in Dhaka, Bangladesh
ID: 009892e02bc1a4c9abf6f547b979e68ecbde8087, Title: Viral respiratory tract infections in young children with cystic fibrosis: a prospective full-year seasonal study
###Markdown
Let's also take a look at the contents of one article
###Code
art = articles[3]
art.id
art.title
art.abstract
art.summary
art.text
###Output
_____no_output_____
###Markdown
Load the pre-trained models
###Code
embedder1 = SentenceTransformer('stsb-distilbert-base')
embedder2 = SentenceTransformer('stsb-roberta-base')
###Output
_____no_output_____
###Markdown
Define Look-up dictionaries to avoid recomputing some embeddings Get the embeddings of the summaries of all the articles, for each model
###Code
# should take ~ 15 mins on GPU for the whole Dataset
index_to_summary_embeddings1 = {}
for idx, article in enumerate(articles):
index_to_summary_embeddings1[idx] = embedder1.encode(article.summary, convert_to_tensor=True)
# should take ~ 27 mins on GPU for the whole Dataset
index_to_summary_embeddings2 = {}
for idx, article in enumerate(articles):
index_to_summary_embeddings2[idx] = embedder2.encode(article.summary, convert_to_tensor=True)
###Output
_____no_output_____
###Markdown
Also define dictionaries that will be used to store the embeddings of the whole articles, to avoid computing them twice
###Code
index_to_article_embedding1 = {}
index_to_article_embedding2 = {}
###Output
_____no_output_____
###Markdown
Predict
###Code
def max_similarity(corpus_embeddings, query_embedding, k=1):
""" returns the indices and values (cosine similarity) of the sentences from the corpus with the
top k cosine similarities with the query embedding """
cos_scores = util.pytorch_cos_sim(query_embedding, corpus_embeddings)[0].cpu()
return torch.topk(cos_scores, k=k)
def relative_articles(articles, index_to_summary_embeddings, query_embedding, threshold=0.5):
""" returns a list with the indices of the articles, for which a sentence in its summary with
cosine similarity > threshold, exists """
return [idx for idx, article in enumerate(articles)
if max_similarity(index_to_summary_embeddings[idx], query_embedding)[0][0] > threshold]
def get_passage(corpus, best_sentence_idx, index_in_section, paragraph, embedder):
""" given a sentence inside the corpus, it returns a string containing all the similar sentences
that belong in the same section of the corpus """
best_sentence_embedding = embedder.encode([corpus[best_sentence_idx]], convert_to_tensor=True)
previous_sentence_is_relevant = True
next_sentence_is_relevant = True
passage = corpus[best_sentence_idx]
offset = 1
# while either the previous sentence of the next are relevant, keep appending them in the passage
while previous_sentence_is_relevant is True or next_sentence_is_relevant is True:
if index_in_section - offset < 0:
previous_sentence_is_relevant = False
elif index_in_section + offset >= len(paragraph):
next_sentence_is_relevant = False
if previous_sentence_is_relevant is True:
previous_sentence = corpus[best_sentence_idx - offset]
previous_sentence_embedding = embedder.encode([previous_sentence], convert_to_tensor=True)
cosine_similarity = util.pytorch_cos_sim(best_sentence_embedding, previous_sentence_embedding)[0][0].cpu()
if cosine_similarity < 0.5:
previous_sentence_is_relevant = False
else:
passage = ' '.join([previous_sentence, passage])
if next_sentence_is_relevant is True:
next_sentence = corpus[best_sentence_idx + offset]
next_sentence_embedding = embedder.encode([next_sentence], convert_to_tensor=True)
cosine_similarity = util.pytorch_cos_sim(best_sentence_embedding, next_sentence_embedding)[0][0].cpu()
if cosine_similarity < 0.5:
next_sentence_is_relevant = False
else:
passage = ' '.join([passage, next_sentence])
offset += 1
return passage
# deprecated
def threshold_is_ok(number_of_articles_included, min_articles=5, max_articles=150):
""" returns a boolean that determines whether the number of articles that did not get filtered
is acceptable """
return min_articles <= number_of_articles_included <= max_articles
# deprecated
def fix_threshold(number_of_articles_included, threshold, min_articles=5, max_articles=150,
decrease_step=0.01, increase_step=0.05):
""" computes a new value for a threshold, depending on how many articles were found relevant
using the previous threshold """
if number_of_articles_included < min_articles:
return threshold + increase_step
elif number_of_articles_included > max_articles:
return threshold - decrease_step
def find_best_article(articles, index_to_summary_embeddings, index_to_article_embeddings, query, embedder, threshold=0.5):
""" returns the index (in the articles list) of the article that best fits the given query """
query_embedding = embedder.encode(query, convert_to_tensor=True)
articles_to_explore = relative_articles(articles, index_to_summary_embeddings, query_embedding, threshold=threshold)
while len(articles_to_explore) == 0:
threshold -= 0.05
articles_to_explore = relative_articles(articles, index_to_summary_embeddings, query_embedding, threshold=threshold)
best_cos_sim = 0.0
best_article_idx = None
best_sentence_idx = None
# print("For query '{}', found {} articles to explore.\n".format(query, len(articles_to_explore)))
# print([articles[article].id for article in articles_to_explore])
for idx in articles_to_explore:
article = articles[idx]
if idx not in index_to_article_embeddings:
corpus = article.text
corpus_embeddings = embedder.encode(corpus, convert_to_tensor=True)
index_to_article_embeddings[idx] = corpus_embeddings
else:
corpus_embeddings = index_to_article_embeddings[idx]
top_sentences_and_scores = max_similarity(corpus_embeddings, query_embedding)
score = top_sentences_and_scores[0][0]
if score > best_cos_sim:
best_cos_sim = score
best_article_idx = idx
best_sentence_idx = top_sentences_and_scores[1][0]
best_article = articles[best_article_idx]
corpus = best_article.text
index_in_section, section = best_article.get_section_from_index(best_sentence_idx)
paragraph = [section[0], *section[1]]
passage = get_passage(corpus, best_sentence_idx, index_in_section, paragraph, embedder)
return best_article_idx, passage
###Output
_____no_output_____
###Markdown
These queries and their respective answers are written in the file queries.txt
###Code
queries = ['How is the diagnosis of pulmonary tuberculosis made in TB clinics and hospitals?',
'Was the Porcine epidemic diarrhea virus first detected in Slovenia?',
'Is handwashing the most important measure against infectious diseases?',
'How was the importance of clathrin-mediated endocytosis for MHV confirmed?',
'Which alternatives do we have for conventional chemotherapy?',
'Is there a drug to treat the EV71 infection?',
'Which is common cause for diarrhea and septicemia in calves?',
'Which scanning technique was used to confirm hypotheses regarding the MAb-1G10 epitope structure?',
'How can host translational inhibition be achieved?',
'What is Multiple sclerosis?']
###Output
_____no_output_____
###Markdown
Define a threshold that will be use to filter out articles where their summary has an a cosine similarity with the query, lower than the threshold. The bigger the threshold, the more articles get filtered out. This has 2 effects:- Greatly improves running time- Increased probability of missing out the best article as it's summary might not be similar to the query, yet its body text may have the correct passage
###Code
threshold = 0.65
###Output
_____no_output_____
###Markdown
Find the best articles for them
###Code
total_time = 0.0
for idx, query in enumerate(queries):
start_time = time.time()
best_article_index, passage = find_best_article(articles, index_to_summary_embeddings1, index_to_article_embedding1, query, embedder1, threshold=threshold)
elapsed_time = time.time() - start_time
total_time += elapsed_time
best_article = articles[best_article_index]
print('-' * 200)
print("For Query {}: '{}', the most relevant article is:\n".format(idx + 1, query))
print('ID: {},\tTitle: {}\n'.format(best_article.id, best_article.title))
print('The Passage is:')
print(passage)
print('\nFound the answer in %.2f seconds' % elapsed_time)
print('-' * 200)
print('\n')
print('\n\n')
print('-' * 200)
print('\nAverage time per Query: {:.2f}\n'.format(total_time / len(queries)))
print('-' * 200)
total_time = 0.0
for idx, query in enumerate(queries):
start_time = time.time()
best_article_index, passage = find_best_article(articles, index_to_summary_embeddings2, index_to_article_embedding2, query, embedder2, threshold=threshold)
elapsed_time = time.time() - start_time
total_time += elapsed_time
best_article = articles[best_article_index]
print('-' * 200)
print("For Query {}: '{}', the most relevant article is:\n".format(idx + 1, query))
print('ID: {},\tTitle: {}\n'.format(best_article.id, best_article.title))
print('The Passage is:')
print(passage)
print('\nFound the answer in %.2f seconds' % elapsed_time)
print('-' * 200)
print('\n')
print('\n\n')
print('-' * 200)
print('\nAverage time per Query: {:.2f}\n'.format(total_time / len(queries)))
print('-' * 200)
###Output
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 1: 'How is the diagnosis of pulmonary tuberculosis made in TB clinics and hospitals?', the most relevant article is:
ID: 0043d044273b8eb1585d3a66061e9b4e03edc062, Title: Evaluation of the tuberculosis programme in Ningxia Hui Autonomous region, the People's Republic of China: a retrospective case study
The Passage is:
Diagnosis of pulmonary TB in hospitals and TB clinics is made on the basis of clinical examination; chest radiography and sputum smear microscopy and/or sputum culture. Following diagnosis, patients enter the DOTS program which prescribes short-course chemotherapy (SCC) comprising 2 months of isoniazid (H), rifampicin (R), pyrazinamide (Z) plus streptomycin (S) or ethambutol (E) followed by 4 months of H and R. This is the WHO recommended regimen for treating new cases of smear-positive pulmonary TB or smear-negative pulmonary TB with substantial radiographic evidence of active disease .
Found the answer in 47.52 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 2: 'Was the Porcine epidemic diarrhea virus first detected in Slovenia?', the most relevant article is:
ID: 105268027d44ab275991e358674462f77223e882, Title: Complete Genome Sequence of the Porcine Epidemic Diarrhea Virus Strain SLO/JH-11/2015
The Passage is:
Porcine epidemic diarrhea virus (PEDV) was detected for the first time in Slovenia in January 2015. The complete genome sequence of PEDV strain SLO/JH-11/2015, obtained from a fecal sample of a fattening pig with diarrhea in September 2015, is closely related to recently detected European strains.
Found the answer in 2.29 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 3: 'Is handwashing the most important measure against infectious diseases?', the most relevant article is:
ID: 1edab5890fbff22ad353739d3d1e80a86d482820, Title: Open Access Handwashing with soap and national handwashing projects in Korea: focus on the National Handwashing Survey, 2006-2014
The Passage is:
Handwashing is the most fundamental way to prevent the spread of infectious diseases. Correct handwashing can prevent 50 to 70% of water-infections and foodborne-infections.
Found the answer in 25.42 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 4: 'How was the importance of clathrin-mediated endocytosis for MHV confirmed?', the most relevant article is:
ID: 3339f4bb346bfa3070ae5fc7dc745ef051535b0e, Title: Coronavirus Cell Entry Occurs through the Endo-/ Lysosomal Pathway in a Proteolysis-Dependent Manner
The Passage is:
Clathrin-mediated endocytosis and late endosomal factors are required for MHV fusion. The importance of clathrin-mediated endocytosis and endosome maturation for MHV fusion was confirmed by analysis of endocytosis-affecting agents using the fusion assay.
Found the answer in 55.34 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 5: 'Which alternatives do we have for conventional chemotherapy?', the most relevant article is:
ID: 63a0d9767212d4316c91660dc4eedc8cb3fe527f, Title: Dihydroberberine exhibits synergistic effects with sunitinib on NSCLC NCI-H460 cells by repressing MAP kinase pathways and inflammatory mediators
The Passage is:
Highly effective and attenuated dose schedules are good regimens for drug research and development. Combination chemotherapy is a good strategy in cancer therapy.
Found the answer in 5.79 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 6: 'Is there a drug to treat the EV71 infection?', the most relevant article is:
ID: e447d139f1d046120a293f1219944e82c77bc829, Title: Oblongifolin M, an active compound isolated from a Chinese medical herb Garcinia oblongifolia, potently inhibits enterovirus 71 reproduction through downregulation of ERp57
The Passage is:
There is no effective drug to treat EV71 infection yet.
Found the answer in 29.16 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 7: 'Which is common cause for diarrhea and septicemia in calves?', the most relevant article is:
ID: b7a6a987030c52cc7ecdf49c3933b6cfda488210, Title: A systematic review and meta-analysis of the epidemiology of pathogenic Escherichia coli of calves and the role of calves as reservoirs for human pathogenic E. coli
The Passage is:
Escherichia coli bacteria are the most common causes of diarrhea and septicemia in calves. Moreover, calves form a major reservoir for transmission of pathogenic E. coli to humans.
Found the answer in 52.88 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 8: 'Which scanning technique was used to confirm hypotheses regarding the MAb-1G10 epitope structure?', the most relevant article is:
ID: 76d39ac4634db5a0fcc8cddbefd965c463c0ace0, Title: Decreased Pattern Recognition Receptor Signaling, Interferon-Signature, and Bactericidal/Permeability- Increasing Protein Gene Expression in Cord Blood of Term Low Birth Weight Human Newborns
The Passage is:
RNA labeling and Affymetrix gene chip expression probe array hybridization.
Found the answer in 5.02 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 9: 'How can host translational inhibition be achieved?', the most relevant article is:
ID: 2e32842d3ebcee3ffaf4b6822dbac0f41f82130a, Title: Associated Virus Vectors
The Passage is:
Further studies demonstrated that scAAV1 and scAAV6 also induce cellular UPR in vitro, with AAV1 vectors activating the PERK pathway (3 fold) while AAV6 vectors induced a significant increase on all the three major UPR pathways [6-16 fold]. These data suggest that the type and strength of UPR activation is dependent on the viral capsid. We then examined if transient inhibition of UPR pathways by RNA interference has an effect on AAV transduction. siRNA mediated silencing of PERK and IRE1a had a modest effect on AAV2 and AAV6 mediated gene expression (,1.5-2 fold) in vitro.
Found the answer in 12.19 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
For Query 10: 'What is Multiple sclerosis?', the most relevant article is:
ID: 86fca5af635ee9425e3375140fb48cbe6d429411, Title: Deep Sequencing for the Detection of Virus-Like Sequences in the Brains of Patients with Multiple Sclerosis: Detection of GBV-C in Human Brain
The Passage is:
It's relationship with the underlying disease, primary progressive multiple sclerosis, in this single person is not clear.
Found the answer in 13.49 seconds
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Average time per Query: 24.91
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
|
notebooks/RankMarkov.ipynb | ###Markdown
Full Rank Markov and Geographic Rank Markov **Author: Wei Kang **
###Code
import libpysal as ps
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
import seaborn as sns
import pandas as pd
import geopandas as gpd
###Output
_____no_output_____
###Markdown
Full Rank Markov
###Code
from giddy.markov import FullRank_Markov
income_table = pd.read_csv(ps.examples.get_path("usjoin.csv"))
income_table.head()
pci = income_table[list(map(str,range(1929,2010)))].values
pci
m = FullRank_Markov(pci)
m.ranks
m.transitions
###Output
_____no_output_____
###Markdown
Full rank Markov transition probability matrix
###Code
m.p
###Output
_____no_output_____
###Markdown
Full rank first mean passage times
###Code
m.fmpt
m.sojourn_time
df_fullrank = pd.DataFrame(np.c_[m.p.diagonal(),m.sojourn_time], columns=["Staying Probability","Sojourn Time"], index = np.arange(m.p.shape[0])+1)
df_fullrank.head()
df_fullrank.plot(subplots=True, layout=(1,2), figsize=(15,5))
sns.distplot(m.fmpt.flatten(),kde=False)
###Output
/Users/weikang/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval
###Markdown
Geographic Rank Markov
###Code
from giddy.markov import GeoRank_Markov, Markov, sojourn_time
gm = GeoRank_Markov(pci)
gm.transitions
gm.p
gm.sojourn_time[:10]
gm.sojourn_time
gm.fmpt
income_table["geo_sojourn_time"] = gm.sojourn_time
i = 0
for state in income_table["Name"]:
income_table["geo_fmpt_to_" + state] = gm.fmpt[:,i]
income_table["geo_fmpt_from_" + state] = gm.fmpt[i,:]
i = i + 1
income_table.head()
geo_table = gpd.read_file(ps.examples.get_path('us48.shp'))
# income_table = pd.read_csv(libpysal.examples.get_path("usjoin.csv"))
complete_table = geo_table.merge(income_table,left_on='STATE_NAME',right_on='Name')
complete_table.head()
complete_table.columns
###Output
_____no_output_____
###Markdown
Visualizing first mean passage time from/to California/Mississippi:
###Code
fig, axes = plt.subplots(nrows=2, ncols=2,figsize = (15,7))
target_states = ["California","Mississippi"]
directions = ["from","to"]
for i, direction in enumerate(directions):
for j, target in enumerate(target_states):
ax = axes[i,j]
col = direction+"_"+target
complete_table.plot(ax=ax,column = "geo_fmpt_"+ col,cmap='OrRd',
scheme='quantiles', legend=True)
ax.set_title("First Mean Passage Time "+direction+" "+target)
ax.axis('off')
leg = ax.get_legend()
leg.set_bbox_to_anchor((0.8, 0.15, 0.16, 0.2))
plt.tight_layout()
###Output
/Users/weikang/anaconda3/lib/python3.6/site-packages/pysal/__init__.py:65: VisibleDeprecationWarning: PySAL's API will be changed on 2018-12-31. The last release made with this API is version 1.14.4. A preview of the next API version is provided in the `pysal` 2.0 prelease candidate. The API changes and a guide on how to change imports is provided at https://migrating.pysal.org
), VisibleDeprecationWarning)
/Users/weikang/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval
###Markdown
Visualizing sojourn time for each US state:
###Code
fig, axes = plt.subplots(nrows=1, ncols=2,figsize = (15,7))
schemes = ["Quantiles","Equal_Interval"]
for i, scheme in enumerate(schemes):
ax = axes[i]
complete_table.plot(ax=ax,column = "geo_sojourn_time",cmap='OrRd',
scheme=scheme, legend=True)
ax.set_title("Rank Sojourn Time ("+scheme+")")
ax.axis('off')
leg = ax.get_legend()
leg.set_bbox_to_anchor((0.8, 0.15, 0.16, 0.2))
plt.tight_layout()
###Output
/Users/weikang/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:1713: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval
|
Convolutional Neural Networks/week_4/Programming_Assignment/Art_Generation_with_Neural_Style_Transfer_v3a.ipynb | ###Markdown
Deep Learning & Art: Neural Style TransferIn this assignment, you will learn about Neural Style Transfer. This algorithm was created by [Gatys et al. (2015).](https://arxiv.org/abs/1508.06576)**In this assignment, you will:**- Implement the neural style transfer algorithm - Generate novel artistic images using your algorithm Most of the algorithms you've studied optimize a cost function to get a set of parameter values. In Neural Style Transfer, you'll optimize a cost function to get pixel values! Updates If you were working on the notebook before this update...* The current notebook is version "3a".* You can find your original work saved in the notebook with the previous version name ("v2") * To view the file directory, go to the menu "File->Open", and this will open a new tab that shows the file directory. List of updates* Use `pprint.PrettyPrinter` to format printing of the vgg model.* computing content cost: clarified and reformatted instructions, fixed broken links, added additional hints for unrolling.* style matrix: clarify two uses of variable "G" by using different notation for gram matrix.* style cost: use distinct notation for gram matrix, added additional hints.* Grammar and wording updates for clarity.* `model_nn`: added hints.
###Code
import os
import sys
import scipy.io
import scipy.misc
import matplotlib.pyplot as plt
from matplotlib.pyplot import imshow
from PIL import Image
from nst_utils import *
import numpy as np
import tensorflow as tf
import pprint
%matplotlib inline
###Output
_____no_output_____
###Markdown
1 - Problem StatementNeural Style Transfer (NST) is one of the most fun techniques in deep learning. As seen below, it merges two images, namely: a **"content" image (C) and a "style" image (S), to create a "generated" image (G**). The generated image G combines the "content" of the image C with the "style" of image S. In this example, you are going to generate an image of the Louvre museum in Paris (content image C), mixed with a painting by Claude Monet, a leader of the impressionist movement (style image S).Let's see how you can do this. 2 - Transfer LearningNeural Style Transfer (NST) uses a previously trained convolutional network, and builds on top of that. The idea of using a network trained on a different task and applying it to a new task is called transfer learning. Following the [original NST paper](https://arxiv.org/abs/1508.06576), we will use the VGG network. Specifically, we'll use VGG-19, a 19-layer version of the VGG network. This model has already been trained on the very large ImageNet database, and thus has learned to recognize a variety of low level features (at the shallower layers) and high level features (at the deeper layers). Run the following code to load parameters from the VGG model. This may take a few seconds.
###Code
pp = pprint.PrettyPrinter(indent=4)
model = load_vgg_model("pretrained-model/imagenet-vgg-verydeep-19.mat")
pp.pprint(model)
###Output
{ 'avgpool1': <tf.Tensor 'AvgPool:0' shape=(1, 150, 200, 64) dtype=float32>,
'avgpool2': <tf.Tensor 'AvgPool_1:0' shape=(1, 75, 100, 128) dtype=float32>,
'avgpool3': <tf.Tensor 'AvgPool_2:0' shape=(1, 38, 50, 256) dtype=float32>,
'avgpool4': <tf.Tensor 'AvgPool_3:0' shape=(1, 19, 25, 512) dtype=float32>,
'avgpool5': <tf.Tensor 'AvgPool_4:0' shape=(1, 10, 13, 512) dtype=float32>,
'conv1_1': <tf.Tensor 'Relu:0' shape=(1, 300, 400, 64) dtype=float32>,
'conv1_2': <tf.Tensor 'Relu_1:0' shape=(1, 300, 400, 64) dtype=float32>,
'conv2_1': <tf.Tensor 'Relu_2:0' shape=(1, 150, 200, 128) dtype=float32>,
'conv2_2': <tf.Tensor 'Relu_3:0' shape=(1, 150, 200, 128) dtype=float32>,
'conv3_1': <tf.Tensor 'Relu_4:0' shape=(1, 75, 100, 256) dtype=float32>,
'conv3_2': <tf.Tensor 'Relu_5:0' shape=(1, 75, 100, 256) dtype=float32>,
'conv3_3': <tf.Tensor 'Relu_6:0' shape=(1, 75, 100, 256) dtype=float32>,
'conv3_4': <tf.Tensor 'Relu_7:0' shape=(1, 75, 100, 256) dtype=float32>,
'conv4_1': <tf.Tensor 'Relu_8:0' shape=(1, 38, 50, 512) dtype=float32>,
'conv4_2': <tf.Tensor 'Relu_9:0' shape=(1, 38, 50, 512) dtype=float32>,
'conv4_3': <tf.Tensor 'Relu_10:0' shape=(1, 38, 50, 512) dtype=float32>,
'conv4_4': <tf.Tensor 'Relu_11:0' shape=(1, 38, 50, 512) dtype=float32>,
'conv5_1': <tf.Tensor 'Relu_12:0' shape=(1, 19, 25, 512) dtype=float32>,
'conv5_2': <tf.Tensor 'Relu_13:0' shape=(1, 19, 25, 512) dtype=float32>,
'conv5_3': <tf.Tensor 'Relu_14:0' shape=(1, 19, 25, 512) dtype=float32>,
'conv5_4': <tf.Tensor 'Relu_15:0' shape=(1, 19, 25, 512) dtype=float32>,
'input': <tf.Variable 'Variable:0' shape=(1, 300, 400, 3) dtype=float32_ref>}
###Markdown
* The model is stored in a python dictionary. * The python dictionary contains key-value pairs for each layer. * The 'key' is the variable name and the 'value' is a tensor for that layer. Assign input image to the model's input layerTo run an image through this network, you just have to feed the image to the model. In TensorFlow, you can do so using the [tf.assign](https://www.tensorflow.org/api_docs/python/tf/assign) function. In particular, you will use the assign function like this: ```pythonmodel["input"].assign(image)```This assigns the image as an input to the model. Activate a layerAfter this, if you want to access the activations of a particular layer, say layer `4_2` when the network is run on this image, you would run a TensorFlow session on the correct tensor `conv4_2`, as follows: ```pythonsess.run(model["conv4_2"])``` 3 - Neural Style Transfer (NST)We will build the Neural Style Transfer (NST) algorithm in three steps:- Build the content cost function $J_{content}(C,G)$- Build the style cost function $J_{style}(S,G)$- Put it together to get $J(G) = \alpha J_{content}(C,G) + \beta J_{style}(S,G)$. 3.1 - Computing the content costIn our running example, the content image C will be the picture of the Louvre Museum in Paris. Run the code below to see a picture of the Louvre.
###Code
content_image = scipy.misc.imread("images/louvre.jpg")
imshow(content_image);
###Output
_____no_output_____
###Markdown
The content image (C) shows the Louvre museum's pyramid surrounded by old Paris buildings, against a sunny sky with a few clouds.** 3.1.1 - Make generated image G match the content of image C** Shallower versus deeper layers* The shallower layers of a ConvNet tend to detect lower-level features such as edges and simple textures.* The deeper layers tend to detect higher-level features such as more complex textures as well as object classes. Choose a "middle" activation layer $a^{[l]}$We would like the "generated" image G to have similar content as the input image C. Suppose you have chosen some layer's activations to represent the content of an image. * In practice, you'll get the most visually pleasing results if you choose a layer in the **middle** of the network--neither too shallow nor too deep. * (After you have finished this exercise, feel free to come back and experiment with using different layers, to see how the results vary.) Forward propagate image "C"* Set the image C as the input to the pretrained VGG network, and run forward propagation. * Let $a^{(C)}$ be the hidden layer activations in the layer you had chosen. (In lecture, we had written this as $a^{[l](C)}$, but here we'll drop the superscript $[l]$ to simplify the notation.) This will be an $n_H \times n_W \times n_C$ tensor. Forward propagate image "G"* Repeat this process with the image G: Set G as the input, and run forward progation. * Let $a^{(G)}$ be the corresponding hidden layer activation. Content Cost Function $J_{content}(C,G)$We will define the content cost function as:$$J_{content}(C,G) = \frac{1}{4 \times n_H \times n_W \times n_C}\sum _{ \text{all entries}} (a^{(C)} - a^{(G)})^2\tag{1} $$* Here, $n_H, n_W$ and $n_C$ are the height, width and number of channels of the hidden layer you have chosen, and appear in a normalization term in the cost. * For clarity, note that $a^{(C)}$ and $a^{(G)}$ are the 3D volumes corresponding to a hidden layer's activations. * In order to compute the cost $J_{content}(C,G)$, it might also be convenient to unroll these 3D volumes into a 2D matrix, as shown below.* Technically this unrolling step isn't needed to compute $J_{content}$, but it will be good practice for when you do need to carry out a similar operation later for computing the style cost $J_{style}$. **Exercise:** Compute the "content cost" using TensorFlow. **Instructions**: The 3 steps to implement this function are:1. Retrieve dimensions from `a_G`: - To retrieve dimensions from a tensor `X`, use: `X.get_shape().as_list()`2. Unroll `a_C` and `a_G` as explained in the picture above - You'll likey want to use these functions: [tf.transpose](https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/transpose) and [tf.reshape](https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/reshape).3. Compute the content cost: - You'll likely want to use these functions: [tf.reduce_sum](https://www.tensorflow.org/api_docs/python/tf/reduce_sum), [tf.square](https://www.tensorflow.org/api_docs/python/tf/square) and [tf.subtract](https://www.tensorflow.org/api_docs/python/tf/subtract). Additional Hints for "Unrolling"* To unroll the tensor, we want the shape to change from $(m,n_H,n_W,n_C)$ to $(m, n_H \times n_W, n_C)$.* `tf.reshape(tensor, shape)` takes a list of integers that represent the desired output shape.* For the `shape` parameter, a `-1` tells the function to choose the correct dimension size so that the output tensor still contains all the values of the original tensor.* So tf.reshape(a_C, shape=[m, n_H * n_W, n_C]) gives the same result as tf.reshape(a_C, shape=[m, -1, n_C]).* If you prefer to re-order the dimensions, you can use `tf.transpose(tensor, perm)`, where `perm` is a list of integers containing the original index of the dimensions. * For example, `tf.transpose(a_C, perm=[0,3,1,2])` changes the dimensions from $(m, n_H, n_W, n_C)$ to $(m, n_C, n_H, n_W)$.* There is more than one way to unroll the tensors.* Notice that it's not necessary to use tf.transpose to 'unroll' the tensors in this case but this is a useful function to practice and understand for other situations that you'll encounter.
###Code
# GRADED FUNCTION: compute_content_cost
def compute_content_cost(a_C, a_G):
"""
Computes the content cost
Arguments:
a_C -- tensor of dimension (1, n_H, n_W, n_C), hidden layer activations representing content of the image C
a_G -- tensor of dimension (1, n_H, n_W, n_C), hidden layer activations representing content of the image G
Returns:
J_content -- scalar that you compute using equation 1 above.
"""
### START CODE HERE ###
# Retrieve dimensions from a_G (≈1 line)
m, n_H, n_W, n_C = a_G.get_shape().as_list()
# Reshape a_C and a_G (≈2 lines)
a_C_unrolled = tf.reshape(a_C,shape=[m,n_H*n_W,n_C])
a_G_unrolled = tf.reshape(a_G,shape=[m,n_H*n_W,n_C])
# compute the cost with tensorflow (≈1 line)
J_content = tf.reduce_sum(tf.square(tf.subtract(a_C_unrolled,a_G_unrolled)))/(4 * n_H * n_W * n_C)
### END CODE HERE ###
return J_content
tf.reset_default_graph()
with tf.Session() as test:
tf.set_random_seed(1)
a_C = tf.random_normal([1, 4, 4, 3], mean=1, stddev=4)
a_G = tf.random_normal([1, 4, 4, 3], mean=1, stddev=4)
J_content = compute_content_cost(a_C, a_G)
print("J_content = " + str(J_content.eval()))
###Output
J_content = 6.76559
###Markdown
**Expected Output**: **J_content** 6.76559 What you should remember- The content cost takes a hidden layer activation of the neural network, and measures how different $a^{(C)}$ and $a^{(G)}$ are. - When we minimize the content cost later, this will help make sure $G$ has similar content as $C$. 3.2 - Computing the style costFor our running example, we will use the following style image:
###Code
style_image = scipy.misc.imread("images/monet_800600.jpg")
imshow(style_image);
###Output
_____no_output_____
###Markdown
This was painted in the style of *[impressionism](https://en.wikipedia.org/wiki/Impressionism)*.Lets see how you can now define a "style" cost function $J_{style}(S,G)$. 3.2.1 - Style matrix Gram matrix* The style matrix is also called a "Gram matrix." * In linear algebra, the Gram matrix G of a set of vectors $(v_{1},\dots ,v_{n})$ is the matrix of dot products, whose entries are ${\displaystyle G_{ij} = v_{i}^T v_{j} = np.dot(v_{i}, v_{j}) }$. * In other words, $G_{ij}$ compares how similar $v_i$ is to $v_j$: If they are highly similar, you would expect them to have a large dot product, and thus for $G_{ij}$ to be large. Two meanings of the variable $G$* Note that there is an unfortunate collision in the variable names used here. We are following common terminology used in the literature. * $G$ is used to denote the Style matrix (or Gram matrix) * $G$ also denotes the generated image. * For this assignment, we will use $G_{gram}$ to refer to the Gram matrix, and $G$ to denote the generated image. Compute $G_{gram}$In Neural Style Transfer (NST), you can compute the Style matrix by multiplying the "unrolled" filter matrix with its transpose:$$\mathbf{G}_{gram} = \mathbf{A}_{unrolled} \mathbf{A}_{unrolled}^T$$ $G_{(gram)i,j}$: correlationThe result is a matrix of dimension $(n_C,n_C)$ where $n_C$ is the number of filters (channels). The value $G_{(gram)i,j}$ measures how similar the activations of filter $i$ are to the activations of filter $j$. $G_{(gram),i,i}$: prevalence of patterns or textures* The diagonal elements $G_{(gram)ii}$ measure how "active" a filter $i$ is. * For example, suppose filter $i$ is detecting vertical textures in the image. Then $G_{(gram)ii}$ measures how common vertical textures are in the image as a whole.* If $G_{(gram)ii}$ is large, this means that the image has a lot of vertical texture. By capturing the prevalence of different types of features ($G_{(gram)ii}$), as well as how much different features occur together ($G_{(gram)ij}$), the Style matrix $G_{gram}$ measures the style of an image. **Exercise**:* Using TensorFlow, implement a function that computes the Gram matrix of a matrix A. * The formula is: The gram matrix of A is $G_A = AA^T$. * You may use these functions: [matmul](https://www.tensorflow.org/api_docs/python/tf/matmul) and [transpose](https://www.tensorflow.org/api_docs/python/tf/transpose).
###Code
# GRADED FUNCTION: gram_matrix
def gram_matrix(A):
"""
Argument:
A -- matrix of shape (n_C, n_H*n_W)
Returns:
GA -- Gram matrix of A, of shape (n_C, n_C)
"""
### START CODE HERE ### (≈1 line)
GA = tf.matmul(A, tf.transpose(A))
### END CODE HERE ###
return GA
tf.reset_default_graph()
with tf.Session() as test:
tf.set_random_seed(1)
A = tf.random_normal([3, 2*1], mean=1, stddev=4)
GA = gram_matrix(A)
print("GA = \n" + str(GA.eval()))
###Output
GA =
[[ 6.42230511 -4.42912197 -2.09668207]
[ -4.42912197 19.46583748 19.56387138]
[ -2.09668207 19.56387138 20.6864624 ]]
###Markdown
**Expected Output**: **GA** [[ 6.42230511 -4.42912197 -2.09668207] [ -4.42912197 19.46583748 19.56387138] [ -2.09668207 19.56387138 20.6864624 ]] 3.2.2 - Style cost Your goal will be to minimize the distance between the Gram matrix of the "style" image S and the gram matrix of the "generated" image G. * For now, we are using only a single hidden layer $a^{[l]}$. * The corresponding style cost for this layer is defined as: $$J_{style}^{[l]}(S,G) = \frac{1}{4 \times {n_C}^2 \times (n_H \times n_W)^2} \sum _{i=1}^{n_C}\sum_{j=1}^{n_C}(G^{(S)}_{(gram)i,j} - G^{(G)}_{(gram)i,j})^2\tag{2} $$* $G_{gram}^{(S)}$ Gram matrix of the "style" image.* $G_{gram}^{(G)}$ Gram matrix of the "generated" image.* Remember, this cost is computed using the hidden layer activations for a particular hidden layer in the network $a^{[l]}$ **Exercise**: Compute the style cost for a single layer. **Instructions**: The 3 steps to implement this function are:1. Retrieve dimensions from the hidden layer activations a_G: - To retrieve dimensions from a tensor X, use: `X.get_shape().as_list()`2. Unroll the hidden layer activations a_S and a_G into 2D matrices, as explained in the picture above (see the images in the sections "computing the content cost" and "style matrix"). - You may use [tf.transpose](https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/transpose) and [tf.reshape](https://www.tensorflow.org/versions/r1.15/api_docs/python/tf/reshape).3. Compute the Style matrix of the images S and G. (Use the function you had previously written.) 4. Compute the Style cost: - You may find [tf.reduce_sum](https://www.tensorflow.org/api_docs/python/tf/reduce_sum), [tf.square](https://www.tensorflow.org/api_docs/python/tf/square) and [tf.subtract](https://www.tensorflow.org/api_docs/python/tf/subtract) useful. Additional Hints* Since the activation dimensions are $(m, n_H, n_W, n_C)$ whereas the desired unrolled matrix shape is $(n_C, n_H*n_W)$, the order of the filter dimension $n_C$ is changed. So `tf.transpose` can be used to change the order of the filter dimension.* for the product $\mathbf{G}_{gram} = \mathbf{A}_{} \mathbf{A}_{}^T$, you will also need to specify the `perm` parameter for the `tf.transpose` function.
###Code
# GRADED FUNCTION: compute_layer_style_cost
def compute_layer_style_cost(a_S, a_G):
"""
Arguments:
a_S -- tensor of dimension (1, n_H, n_W, n_C), hidden layer activations representing style of the image S
a_G -- tensor of dimension (1, n_H, n_W, n_C), hidden layer activations representing style of the image G
Returns:
J_style_layer -- tensor representing a scalar value, style cost defined above by equation (2)
"""
### START CODE HERE ###
# Retrieve dimensions from a_G (≈1 line)
m, n_H, n_W, n_C = a_G.get_shape().as_list()
# Reshape the images to have them of shape (n_C, n_H*n_W) (≈2 lines)
a_S = tf.reshape(tf.transpose(a_S, perm = [0,3,1,2]),shape=[n_C,n_H * n_W])
a_G = tf.reshape(tf.transpose(a_G, perm= [0,3,1,2]),shape=[n_C,n_H * n_W])
# Computing gram_matrices for both images S and G (≈2 lines)
GS = gram_matrix(a_S)
GG = gram_matrix(a_G)
# Computing the loss (≈1 line)
J_style_layer = tf.reduce_sum(tf.square(tf.subtract(GS,GG)))/(4 * n_C**2 * (n_H * n_W)**2)
### END CODE HERE ###
return J_style_layer
tf.reset_default_graph()
with tf.Session() as test:
tf.set_random_seed(1)
a_S = tf.random_normal([1, 4, 4, 3], mean=1, stddev=4)
a_G = tf.random_normal([1, 4, 4, 3], mean=1, stddev=4)
J_style_layer = compute_layer_style_cost(a_S, a_G)
print("J_style_layer = " + str(J_style_layer.eval()))
###Output
J_style_layer = 9.19028
###Markdown
**Expected Output**: **J_style_layer** 9.19028 3.2.3 Style Weights* So far you have captured the style from only one layer. * We'll get better results if we "merge" style costs from several different layers. * Each layer will be given weights ($\lambda^{[l]}$) that reflect how much each layer will contribute to the style.* After completing this exercise, feel free to come back and experiment with different weights to see how it changes the generated image $G$.* By default, we'll give each layer equal weight, and the weights add up to 1. ($\sum_{l}^L\lambda^{[l]} = 1$)
###Code
STYLE_LAYERS = [
('conv1_1', 0.2),
('conv2_1', 0.2),
('conv3_1', 0.2),
('conv4_1', 0.2),
('conv5_1', 0.2)]
###Output
_____no_output_____
###Markdown
You can combine the style costs for different layers as follows:$$J_{style}(S,G) = \sum_{l} \lambda^{[l]} J^{[l]}_{style}(S,G)$$where the values for $\lambda^{[l]}$ are given in `STYLE_LAYERS`. Exercise: compute style cost* We've implemented a compute_style_cost(...) function. * It calls your `compute_layer_style_cost(...)` several times, and weights their results using the values in `STYLE_LAYERS`. * Please read over it to make sure you understand what it's doing. Description of `compute_style_cost`For each layer:* Select the activation (the output tensor) of the current layer.* Get the style of the style image "S" from the current layer.* Get the style of the generated image "G" from the current layer.* Compute the "style cost" for the current layer* Add the weighted style cost to the overall style cost (J_style)Once you're done with the loop: * Return the overall style cost.
###Code
def compute_style_cost(model, STYLE_LAYERS):
"""
Computes the overall style cost from several chosen layers
Arguments:
model -- our tensorflow model
STYLE_LAYERS -- A python list containing:
- the names of the layers we would like to extract style from
- a coefficient for each of them
Returns:
J_style -- tensor representing a scalar value, style cost defined above by equation (2)
"""
# initialize the overall style cost
J_style = 0
for layer_name, coeff in STYLE_LAYERS:
# Select the output tensor of the currently selected layer
out = model[layer_name]
# Set a_S to be the hidden layer activation from the layer we have selected, by running the session on out
a_S = sess.run(out)
# Set a_G to be the hidden layer activation from same layer. Here, a_G references model[layer_name]
# and isn't evaluated yet. Later in the code, we'll assign the image G as the model input, so that
# when we run the session, this will be the activations drawn from the appropriate layer, with G as input.
a_G = out
# Compute style_cost for the current layer
J_style_layer = compute_layer_style_cost(a_S, a_G)
# Add coeff * J_style_layer of this layer to overall style cost
J_style += coeff * J_style_layer
return J_style
###Output
_____no_output_____
###Markdown
**Note**: In the inner-loop of the for-loop above, `a_G` is a tensor and hasn't been evaluated yet. It will be evaluated and updated at each iteration when we run the TensorFlow graph in model_nn() below.<!-- How do you choose the coefficients for each layer? The deeper layers capture higher-level concepts, and the features in the deeper layers are less localized in the image relative to each other. So if you want the generated image to softly follow the style image, try choosing larger weights for deeper layers and smaller weights for the first layers. In contrast, if you want the generated image to strongly follow the style image, try choosing smaller weights for deeper layers and larger weights for the first layers!--> What you should remember- The style of an image can be represented using the Gram matrix of a hidden layer's activations. - We get even better results by combining this representation from multiple different layers. - This is in contrast to the content representation, where usually using just a single hidden layer is sufficient.- Minimizing the style cost will cause the image $G$ to follow the style of the image $S$. 3.3 - Defining the total cost to optimize Finally, let's create a cost function that minimizes both the style and the content cost. The formula is: $$J(G) = \alpha J_{content}(C,G) + \beta J_{style}(S,G)$$**Exercise**: Implement the total cost function which includes both the content cost and the style cost.
###Code
# GRADED FUNCTION: total_cost
def total_cost(J_content, J_style, alpha = 10, beta = 40):
"""
Computes the total cost function
Arguments:
J_content -- content cost coded above
J_style -- style cost coded above
alpha -- hyperparameter weighting the importance of the content cost
beta -- hyperparameter weighting the importance of the style cost
Returns:
J -- total cost as defined by the formula above.
"""
### START CODE HERE ### (≈1 line)
J = alpha*J_content + beta*J_style
### END CODE HERE ###
return J
tf.reset_default_graph()
with tf.Session() as test:
np.random.seed(3)
J_content = np.random.randn()
J_style = np.random.randn()
J = total_cost(J_content, J_style)
print("J = " + str(J))
###Output
J = 35.34667875478276
###Markdown
**Expected Output**: **J** 35.34667875478276 What you should remember- The total cost is a linear combination of the content cost $J_{content}(C,G)$ and the style cost $J_{style}(S,G)$.- $\alpha$ and $\beta$ are hyperparameters that control the relative weighting between content and style. 4 - Solving the optimization problem Finally, let's put everything together to implement Neural Style Transfer!Here's what the program will have to do:1. Create an Interactive Session2. Load the content image 3. Load the style image4. Randomly initialize the image to be generated 5. Load the VGG19 model7. Build the TensorFlow graph: - Run the content image through the VGG19 model and compute the content cost - Run the style image through the VGG19 model and compute the style cost - Compute the total cost - Define the optimizer and the learning rate8. Initialize the TensorFlow graph and run it for a large number of iterations, updating the generated image at every step.Lets go through the individual steps in detail. Interactive SessionsYou've previously implemented the overall cost $J(G)$. We'll now set up TensorFlow to optimize this with respect to $G$. * To do so, your program has to reset the graph and use an "[Interactive Session](https://www.tensorflow.org/api_docs/python/tf/InteractiveSession)". * Unlike a regular session, the "Interactive Session" installs itself as the default session to build a graph. * This allows you to run variables without constantly needing to refer to the session object (calling "sess.run()"), which simplifies the code. Start the interactive session.
###Code
# Reset the graph
tf.reset_default_graph()
# Start interactive session
sess = tf.InteractiveSession()
###Output
_____no_output_____
###Markdown
Content imageLet's load, reshape, and normalize our "content" image (the Louvre museum picture):
###Code
content_image = scipy.misc.imread("images/louvre_small.jpg")
content_image = reshape_and_normalize_image(content_image)
###Output
_____no_output_____
###Markdown
Style imageLet's load, reshape and normalize our "style" image (Claude Monet's painting):
###Code
style_image = scipy.misc.imread("images/monet.jpg")
style_image = reshape_and_normalize_image(style_image)
###Output
_____no_output_____
###Markdown
Generated image correlated with content imageNow, we initialize the "generated" image as a noisy image created from the content_image.* The generated image is slightly correlated with the content image.* By initializing the pixels of the generated image to be mostly noise but slightly correlated with the content image, this will help the content of the "generated" image more rapidly match the content of the "content" image. * Feel free to look in `nst_utils.py` to see the details of `generate_noise_image(...)`; to do so, click "File-->Open..." at the upper-left corner of this Jupyter notebook.
###Code
generated_image = generate_noise_image(content_image)
imshow(generated_image[0]);
###Output
_____no_output_____
###Markdown
Load pre-trained VGG19 modelNext, as explained in part (2), let's load the VGG19 model.
###Code
model = load_vgg_model("pretrained-model/imagenet-vgg-verydeep-19.mat")
###Output
_____no_output_____
###Markdown
Content CostTo get the program to compute the content cost, we will now assign `a_C` and `a_G` to be the appropriate hidden layer activations. We will use layer `conv4_2` to compute the content cost. The code below does the following:1. Assign the content image to be the input to the VGG model.2. Set a_C to be the tensor giving the hidden layer activation for layer "conv4_2".3. Set a_G to be the tensor giving the hidden layer activation for the same layer. 4. Compute the content cost using a_C and a_G.**Note**: At this point, a_G is a tensor and hasn't been evaluated. It will be evaluated and updated at each iteration when we run the Tensorflow graph in model_nn() below.
###Code
# Assign the content image to be the input of the VGG model.
sess.run(model['input'].assign(content_image))
# Select the output tensor of layer conv4_2
out = model['conv4_2']
# Set a_C to be the hidden layer activation from the layer we have selected
a_C = sess.run(out)
# Set a_G to be the hidden layer activation from same layer. Here, a_G references model['conv4_2']
# and isn't evaluated yet. Later in the code, we'll assign the image G as the model input, so that
# when we run the session, this will be the activations drawn from the appropriate layer, with G as input.
a_G = out
# Compute the content cost
J_content = compute_content_cost(a_C, a_G)
###Output
_____no_output_____
###Markdown
Style cost
###Code
# Assign the input of the model to be the "style" image
sess.run(model['input'].assign(style_image))
# Compute the style cost
J_style = compute_style_cost(model, STYLE_LAYERS)
###Output
_____no_output_____
###Markdown
Exercise: total cost* Now that you have J_content and J_style, compute the total cost J by calling `total_cost()`. * Use `alpha = 10` and `beta = 40`.
###Code
### START CODE HERE ### (1 line)
J = total_cost(J_content, J_style, alpha=10, beta=40)
### END CODE HERE ###
###Output
_____no_output_____
###Markdown
Optimizer* Use the Adam optimizer to minimize the total cost `J`.* Use a learning rate of 2.0. * [Adam Optimizer documentation](https://www.tensorflow.org/api_docs/python/tf/train/AdamOptimizer)
###Code
# define optimizer (1 line)
optimizer = tf.train.AdamOptimizer(2.0)
# define train_step (1 line)
train_step = optimizer.minimize(J)
###Output
_____no_output_____
###Markdown
Exercise: implement the model* Implement the model_nn() function. * The function **initializes** the variables of the tensorflow graph, * **assigns** the input image (initial generated image) as the input of the VGG19 model * and **runs** the `train_step` tensor (it was created in the code above this function) for a large number of steps. Hints* To initialize global variables, use this: ```Pythonsess.run(tf.global_variables_initializer())```* Run `sess.run()` to evaluate a variable.* [assign](https://www.tensorflow.org/versions/r1.14/api_docs/python/tf/assign) can be used like this:```pythonmodel["input"].assign(image)```
###Code
def model_nn(sess, input_image, num_iterations = 200):
# Initialize global variables (you need to run the session on the initializer)
### START CODE HERE ### (1 line)
sess.run(tf.global_variables_initializer())
### END CODE HERE ###
# Run the noisy input image (initial generated image) through the model. Use assign().
### START CODE HERE ### (1 line)
generated_image = sess.run(model["input"].assign(input_image))
### END CODE HERE ###
for i in range(num_iterations):
# Run the session on the train_step to minimize the total cost
### START CODE HERE ### (1 line)
sess.run(train_step)
### END CODE HERE ###
# Compute the generated image by running the session on the current model['input']
### START CODE HERE ### (1 line)
generated_image = sess.run(model["input"])
### END CODE HERE ###
# Print every 20 iteration.
if i%20 == 0:
Jt, Jc, Js = sess.run([J, J_content, J_style])
print("Iteration " + str(i) + " :")
print("total cost = " + str(Jt))
print("content cost = " + str(Jc))
print("style cost = " + str(Js))
# save current generated image in the "/output" directory
save_image("output/" + str(i) + ".png", generated_image)
# save last generated image
save_image('output/generated_image.jpg', generated_image)
return generated_image
###Output
_____no_output_____
###Markdown
Run the following cell to generate an artistic image. It should take about 3min on CPU for every 20 iterations but you start observing attractive results after ≈140 iterations. Neural Style Transfer is generally trained using GPUs.
###Code
model_nn(sess, generated_image)
###Output
Iteration 0 :
total cost = 5.05035e+09
content cost = 7877.68
style cost = 1.26257e+08
Iteration 20 :
total cost = 9.43276e+08
content cost = 15187.0
style cost = 2.35781e+07
Iteration 40 :
total cost = 4.84898e+08
content cost = 16785.0
style cost = 1.21183e+07
Iteration 60 :
total cost = 3.12574e+08
content cost = 17465.8
style cost = 7.80998e+06
Iteration 80 :
total cost = 2.28137e+08
content cost = 17715.0
style cost = 5.699e+06
Iteration 100 :
total cost = 1.80694e+08
content cost = 17895.5
style cost = 4.51288e+06
Iteration 120 :
total cost = 1.49996e+08
content cost = 18034.4
style cost = 3.74539e+06
Iteration 140 :
total cost = 1.27698e+08
content cost = 18186.9
style cost = 3.18791e+06
Iteration 160 :
total cost = 1.10698e+08
content cost = 18354.2
style cost = 2.76287e+06
Iteration 180 :
total cost = 9.73408e+07
content cost = 18501.0
style cost = 2.4289e+06
|
notebooks/automl/02_automated_ML.ipynb | ###Markdown
 Classification using Automated MLIn this example we use Azure ML's Automated ML functionality to improve on the classifier we built earlier. Automated ML handles the task of building many models from a wide variety of algorithms and choosing a good set of hyper-parameters for them. We then select best the model (or one that meets our criteria) and deploy it as a web service. Load and prepare experiment As part of the setup we have already created an AML workspace. Let's load the workspace and create an experiment.
###Code
import json
import logging
import os
import random
from matplotlib import pyplot as plt
from matplotlib.pyplot import imshow
import pandas as pd
from sklearn import datasets
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
from sklearn.metrics import roc_auc_score
import azureml.core
from azureml.core.experiment import Experiment
from azureml.core.workspace import Workspace
from azureml.train.automl import AutoMLConfig
from azureml.train.automl.run import AutoMLRun
###Output
_____no_output_____
###Markdown
We load the workspace directly from the config file we created in the early part of the course.
###Code
config_path = '/dbfs/tmp/aml_config'
ws = Workspace.from_config(path=os.path.join(config_path, 'config.json'))
experiment_name = 'pred-maint-automl' # choose a name for experiment
project_folder = '.' # project folder
experiment=Experiment(ws, experiment_name)
output = {}
output['SDK version'] = azureml.core.VERSION
output['Subscription ID'] = ws.subscription_id
output['Workspace'] = ws.name
output['Resource Group'] = ws.resource_group
output['Location'] = ws.location
output['Project Directory'] = project_folder
output['Experiment Name'] = experiment.name
pd.set_option('display.max_colwidth', -1)
pd.DataFrame(data=output, index=['']).T
###Output
_____no_output_____
###Markdown
Opt in for diagnostics for better experience, quality, and security of future releases:
###Code
from azureml.telemetry import set_diagnostics_collection
set_diagnostics_collection(send_diagnostics=True)
###Output
_____no_output_____
###Markdown
Instantiate config file We now instantiate a `AutoMLConfig` object. This defines the settings and data used to run the experiment.|Property|Description||-|-||**task**|classification or regression||**primary_metric**|This is the metric that you want to optimize. Classification supports the following primary metrics accuracyAUC_weightedbalanced_accuracyaverage_precision_score_weightedprecision_score_weighted||**max_time_sec**|Time limit in seconds for each iterations||**iterations**|Number of iterations. In each iteration Auto ML trains the data with a specific pipeline||**n_cross_validations**|Number of cross validation splits||**X**|(sparse) array-like, shape = [n_samples, n_features]||**y**|(sparse) array-like, shape = [n_samples, ], [n_samples, n_classes]Multi-class targets. An indicator matrix turns on multilabel classification. This should be an array of integers. ||**path**|Relative path to the project folder. AutoML stores configuration files for the experiment under this folder. You can specify a new empty folder. |
###Code
df = spark.read.parquet("dbfs:/FileStore/tables/preprocessed").cache()
display(df)
from pandas import datetime
from pyspark.sql.functions import col, hour
# we sample every nth row of the data using the `hour` function
df_train = df.filter((col('datetime') < datetime(2015, 10, 1))) # & (hour(col('datetime')) % 3 == 0))
df_test = df.filter(col('datetime') > datetime(2015, 10, 15))
X_keep = ['diff_maint_1', 'diff_error_1', 'volt_sd_3', 'diff_fail_3', 'pressure_ma_3', 'pressure_sd_3', 'diff_fail_1', 'diff_fail_0', 'age', 'vibration_ma_3', 'rotate_ma_3', 'diff_error_2', 'diff_fail_2', 'diff_error_3', 'diff_maint_2', 'volt_ma_3', 'diff_maint_0', 'vibration_sd_3', 'diff_maint_3', 'rotate_sd_3', 'diff_error_0', 'diff_error_4']
Y_keep = ['y_1'] # ['y_0', 'y_1', 'y_2', 'y_3']
# for now, we convert the spark DataFrames to Pandas dataframes,
# because at this point automated ML only supports the latter.
# This will change soon though.
X_train = df_train.select(X_keep).toPandas()
X_test = df_test.select(X_keep).toPandas()
y_train = df_train.select(Y_keep).toPandas()
y_test = df_test.select(Y_keep).toPandas()
X_train.columns
###Output
_____no_output_____
###Markdown
Here are the metrics we can choose to optimize our model over.
###Code
azureml.train.automl.constants.Metric.CLASSIFICATION_PRIMARY_SET
###Output
_____no_output_____
###Markdown
Since we are using `automl` on top of Spark, we have an additional step to run before running the experiment: We use the `azureml.dataprep` API to turn the data from a Pandas DataFrame to a Dataflow. A Dataflow represents a series of lazily-evaluated, immutable operations on data.
###Code
%sh
rm -r /dbfs/dprep
mkdir /dbfs/dprep
mkdir /dbfs/dprep/tmp
import azureml.dataprep as dprep
from pyspark.sql.functions import rand
temp_location = '/dbfs/dprep/tmp'
X_dflow = dprep.read_pandas_dataframe(X_train, temp_folder = temp_location + '/X')
y_dflow = dprep.read_pandas_dataframe(y_train, temp_folder = temp_location + '/y')
###Output
_____no_output_____
###Markdown
We now set up a configuration file for the automated ML training experiment. It contains details for how the experiment should run.
###Code
num_iters = 15
automl_config = AutoMLConfig(task = 'classification',
preprocess = False,
name = experiment_name,
debug_log = 'automl_errors.log',
primary_metric = 'AUC_weighted',
iteration_timeout_minutes = 15,
iterations = num_iters,
verbosity = logging.INFO,
spark_context = sc,
X = X_dflow,
y = y_dflow,
# validation_size = 0.10,
n_cross_validations = 3,
path = project_folder)
###Output
_____no_output_____
###Markdown
Run training experiment You can call the submit method on the experiment object and pass the run configuration. For Local runs the execution is synchronous. Depending on the data and number of iterations this can run for while.You will see the currently running iterations printing to the console.
###Code
local_run = experiment.submit(automl_config, show_output = True)
###Output
_____no_output_____
###Markdown
Portal URL for Monitoring RunsThe following will provide a link to the web interface to explore individual run details and status.
###Code
displayHTML("<a href={} target='_blank'>Your experiment in Azure Portal: {}</a>".format(local_run.get_portal_url(), local_run.id))
###Output
_____no_output_____
###Markdown
Retrieve the Best ModelBelow we select the best pipeline from our iterations. The *get_output* method on automl_classifier returns the best run and the fitted model for the last *fit* invocation. There are overloads on *get_output* that allow you to retrieve the best run and fitted model for *any* logged metric or a particular *iteration*.
###Code
best_run, fitted_model = local_run.get_output()
fitted_model
###Output
_____no_output_____
###Markdown
We can see from the above results that `StandardScalerWrapper` was used to scale the features and a `LightGBMClassifier` was chosen as the best model based on the metric we defined. This of course does NOT automatically also make it the best model in production, but choosing the right model for production is beyond the scope of this course so we will not address it here. Hands-on lab Run the following cell and go to the link provided under Details Page. This links will take us to the Azure portal. Examine the content of the page. Can you find what resource group this resource is under? What kind of resource is it?
###Code
displayHTML("<a href={} target='_blank'>Your experiment in Azure Portal: {}</a>".format(best_run.get_portal_url(), best_run.id))
###Output
_____no_output_____
###Markdown
In addition to choosing a good algorithm, the experiment also tuned hyper-parameters. So our model didn't just run with the default hyper-parameter values. Find out how we can get the chosen hyper-parameters from the `fitted_model` object. Describe the hyper-parameters you see. Which ones do you think are the most critical ones?
###Code
# write solution here
###Output
_____no_output_____
###Markdown
End of lab We can run the following code to get the hyper-parameters for the chosen model.
###Code
n_steps = len(fitted_model.steps)
for s in range(n_steps):
print("Step: %s" % s)
params = fitted_model.steps[s][1].get_params()
print(params)
###Output
_____no_output_____
###Markdown
With a little bit of re-arranging we can pull out the metrics we're interested in and compare them across different runs. Since we're in Databricks, we can use `display` to create a quick visualization with the results.
###Code
children = list(local_run.get_children())
metricslist = {}
for run in children:
properties = run.get_properties()
metrics = {k: v for k, v in run.get_metrics().items() if isinstance(v, float)}
metricslist[int(properties['iteration'])] = metrics
rundata = pd.DataFrame(metricslist).sort_index(1)
rundata['metric'] = rundata.index
rundata = rundata.reset_index()
res = rundata.melt(id_vars='metric', value_vars = range(num_iters))
display(res.loc[res.metric.str.contains('micro'), :]) # in this case we limit the metrics to the ones containing the word `micro`
###Output
_____no_output_____
###Markdown
Quick note: macro and micro averages are only relevant in a multi-class classification scenario. In this situation, we can compute a single classification metric (such as precison, recall or AUC) in one of two ways. Let's use precision to illustrate the difference. Recall that precision = TP / (TP + FP). In a multi-class classification setting, each class has a TP and FP, and hence a precision. So we can compute a macro and micro precision as such:- We can compute precision *per class* and then average across all classes. This is called a **macro precision**. When class imbalances are strong macro averages are not recommended, because by taking a *simple* (not weighted) average we treat all classes equally.- We can compute the aggregated value for TP and FS (summed up for all classes) and then compute an aggregated precision. This is called a **micro presicion** and it works better when we have imbalances across the classes. Score and evaluate the chosen model Let's now pick the best model returned by the experiment and use it to get predictions for the test data. This is simply done by replacing `manual_model.predict` with `fitted_model.predict`.
###Code
y_pred = fitted_model.predict(X_test)
###Output
_____no_output_____
###Markdown
We should get the same confusion matrix we did in the section above.
###Code
confusion_matrix(y_test.values, y_pred)
###Output
_____no_output_____
###Markdown
We use `classification_report` to automatically calculate precision, recall, and the F-1 score from the confusion matrix above.
###Code
cl_report = classification_report(y_test.values, y_pred)
print(cl_report)
###Output
_____no_output_____
###Markdown
The AUC is just the area under the ROC curve shown here:
###Code
from sklearn.metrics import auc, roc_curve
fpr, tpr, thresholds = roc_curve(y_test.values, y_pred)
roc_auc = auc(fpr, tpr)
import matplotlib.pyplot as plt
plt.plot(fpr, tpr, 'b', label = 'AUC = {0:.2f}'.format(roc_auc))
plt.legend(loc = 'lower right')
plt.plot([0, 1], [0, 1],'r--')
plt.xlim([0, 1])
plt.ylim([0, 1])
plt.ylabel('True Positive Rate')
plt.xlabel('False Positive Rate')
plt.show()
display()
###Output
_____no_output_____
###Markdown
Register fitted model for deployment Now that we have a model we're happy with, we register it to our Azure ML account. This will be the first step toward model management and deployment, which we cover in the next Notebook. Registered models can also be loaded into other workspaces.
###Code
description = 'automated ML PdM (predict y_1)'
tags = None
model = local_run.register_model(description=description, tags=tags)
local_run.model_id # Use this id to deploy the model as a web service in Azur
###Output
_____no_output_____
###Markdown
Optionally, we can also take the model and save it on disk as a pickle file, as shown here:
###Code
from sklearn.externals import joblib
joblib.dump(value=fitted_model, filename='model.pkl')
# You can ignore this code, we use it for testing our notebooks.
assert isinstance(model, azureml.core.model.Model)
###Output
_____no_output_____ |
10 Days of Statistics/Day_7. Pearson Correlation Coefficient I.ipynb | ###Markdown
Day_7. Pearson Correlation Coefficient I`Task`Given two `n`-element data sets, `X` and , `Y` calculate the value of the Pearson correlation coefficient.`Input Format`The first line contains an integer, `n`, denoting the size of data sets `X` and `Y`. The second line contains `n` space-separated real numbers (scaled to at most one decimal place), defining data set `X`. The third line contains `n` space-separated real numbers (scaled to at most one decimal place), defining data set `Y`.`Constraints`````10 <= n <= 100``1 <= Xi <= 500`, where `Xi` is the `i th` value of data set `X`.`1 <= Yi <= 500`, where `Yi` is the `i th` value of data set `Y`.Data set `X` contains unique values.Data set `Y` contains unique values.````Output Format`Print the value of the Pearson correlation coefficient, rounded to a scale of `3` decimal places.`Sample Input````1010 9.8 8 7.8 7.7 7 6 5 4 2 200 44 32 24 22 17 15 12 8 4````Sample Output````0.612```
###Code
from math import *
n = int(input())
X = tuple(map(float, input().split()))
Y = tuple(map(float, input().split()))
X_bar = sum(X)/len(X)
Y_bar = sum(Y)/len(Y)
X_std = sqrt(sum(list(map(lambda x: x**2, X)))/len(X) - X_bar**2)
Y_std = sqrt(sum(list(map(lambda y: y**2, Y)))/len(Y) - Y_bar**2)
XY_cov = sum([x*y for x, y in zip(X, Y)])/len(X) - X_bar * Y_bar
ro = XY_cov / (X_std * Y_std)
print(f'{ro:.3f}')
###Output
_____no_output_____ |
examples/grids/smib_milano_ex8p1/smib_milano_ex8p1_4ord_avr/smib_milano_ex8p1_4ord_avr_pss.ipynb | ###Markdown
SMIB system as in Milano's book example 8.1
###Code
%matplotlib widget
import numpy as np
import matplotlib.pyplot as plt
import scipy.optimize as sopt
import ipywidgets
from pydae import ssa
import json
###Output
_____no_output_____
###Markdown
Import system module
###Code
from smib_milano_ex8p1_4ord_avr_pss import smib_milano_ex8p1_4ord_avr_pss_class
###Output
_____no_output_____
###Markdown
Instantiate system
###Code
smib = smib_milano_ex8p1_4ord_avr_pss_class()
xy_0_dict = {
'omega':1,'v_ref':1,'v_c':1
}
###Output
_____no_output_____
###Markdown
Initialize the system (backward and foreward)
###Code
events=[{'p_m':0.8, 'v_t':1.0, 'K_a':100, 'T_e':0.1}]
smib.initialize(events,xy_0_dict)
smib.save_0()
smib.ss()
smib.report_u()
smib.report_x()
smib.report_y()
ssa.eval_A(smib)
ssa.damp_report(smib)
smib = smib_milano_ex8p1_4ord_avr_pss_class()
smib.load_0('xy_0.json')
smib.ss()
smib.eval_jacobians()
smib.eval_A()
ssa.damp_report(smib)
smib.report_params()
smib.load_0('xy_0.json')
def obj(x):
T_1 = x[0]
#K_stab = x[1]
smib.set_value('T_1',T_1)
freq = 1.2
T_2 = 0.1
cplx = (1j*2*np.pi*freq*T_1 + 1)/(1j*2*np.pi*freq*T_2 + 1)
smib.set_value('K_stab',1)
smib.set_value('K_a',100)
smib.set_value('H',6)
smib.ss()
smib.eval_jacobians()
ssa.eval_A(smib)
eig_values,eig_vectors = np.linalg.eig(smib.A)
zetas = -eig_values.real/np.abs(eig_values)
return -100*np.min(zetas)
sopt.differential_evolution(obj,bounds=[(0.1,5)])
events=[{'t_end':1.0},
{'t_end':5.0, 'v_ref':v_ref_0*1.05}
]
smib.simulate(events,xy0='prev');
plt.close('all')
fig, axes = plt.subplots(nrows=2,ncols=2, figsize=(10, 5), frameon=False, dpi=70)
axes[0,0].plot(smib.T, smib.get_values('omega'), label=f'$\omega$')
axes[0,1].plot(smib.T, smib.get_values('v_t'), label=f'$v_t$')
axes[1,0].plot(smib.T, smib.get_values('p_t'), label=f'$p_t$')
axes[1,1].plot(smib.T, smib.get_values('q_t'), label=f'$q_t$')
for ax in axes.flatten():
ax.legend()
###Output
_____no_output_____
###Markdown
Simulation
###Code
smib = smib_milano_ex8p1_4ord_avr_pss_class()
events=[{'p_t':0.8, 'v_t':1.0, 'K_a':200, 'T_e':0.2, 'H':6, 'K_stab':0, 'T_1':0.1}]
smib.initialize(events,xy0=1)
v_ref_0 = smib.get_value('v_ref')
events=[{'t_end':1.0},
{'t_end':15.0, 'v_ref':v_ref_0*1.05}
]
smib.simulate(events,xy0='prev');
plt.close('all')
fig, axes = plt.subplots(nrows=2,ncols=2, figsize=(10, 5), frameon=False, dpi=70)
axes[0,0].plot(smib.T, smib.get_values('omega'), label=f'$\omega$')
axes[0,1].plot(smib.T, smib.get_values('v_t'), label=f'$v_t$')
axes[1,0].plot(smib.T, smib.get_values('p_t'), label=f'$p_t$')
axes[1,1].plot(smib.T, smib.get_values('q_t'), label=f'$q_t$')
for ax in axes.flatten():
ax.legend()
smib = smib_milano_ex8p1_4ord_avr_pss_class()
events=[{'p_t':0.8, 'v_t':1.0, 'K_a':200, 'T_e':0.2, 'H':6, 'K_stab':0, 'T_1':0.1}]
smib.initialize(events,xy_0_dict)
ssa.eval_A(smib)
ssa.damp_report(smib)
###Output
_____no_output_____
###Markdown
Run in two time intervals
###Code
events=[{'t_end':1.0}]
syst.run(events)
events=[{'t_end':2.0}]
syst.run(events)
syst.get_value('omega')
events=[{'p_t':0.8, 'v_t':1.0, 'K_a':100, 'T_e':0.2, 'H':6, 'K_stab':0, 'T_1':0.1}]
smib.initialize(events,xy0=1)
ssa.eval_A(smib)
ssa.damp_report(smib)
ssa.participation(smib).abs().round(2)
smib.report_params()
Ts_control = 0.010
times = np.arange(0.0,10,Ts_control)
# Calculate second references
events=[{'P_t':0.9, 'Q_t':0.0}]
syst.initialize(events,xy0=1.0)
x_ref = np.copy(syst.struct[0].x)
v_f_ref = syst.struct[0]['v_f']
p_m_ref = syst.struct[0]['p_m']
# Calculate initial references
events=[{'P_t':0.0, 'Q_t':0.0}]
syst.initialize(events,xy0=1.0)
x_0 = np.copy(syst.struct[0].x)
v_f_0 = syst.get_value('v_f')
p_m_0 = syst.get_value('p_m')
# Control design
ssa.eval_ss(syst)
Q = np.eye(syst.N_x)*100
R = np.eye(syst.N_u)
K = ctrl.place(syst.A,syst.B,[-2.0+1j*6,-2.0-1j*6,-100,-101])
K,S,E = ctrl.lqr(syst.A,syst.B,Q,R)
Ad,Bd = ssa.discretise_time(syst.A,syst.B,Ts_control)
Kd,S,E = ssa.dlqr(Ad,Bd,Q,R)
for t in times:
x = np.copy(syst.struct[0].x)
v_f = v_f_0
p_m = p_m_0
if t>1.0:
u_ctrl = K*(x_ref - x)
p_m = p_m_ref + u_ctrl[0]
v_f = v_f_ref + u_ctrl[1]
events=[{'t_end':t,'v_f':v_f,'p_m':p_m}]
syst.run(events)
syst.post();
plt.close('all')
fig, axes = plt.subplots(nrows=2,ncols=2, figsize=(10, 5), frameon=False, dpi=50)
axes[0,0].plot(syst.T, syst.get_values('omega'), label=f'$\omega$')
axes[0,1].plot(syst.T, syst.get_values('v_1'), label=f'$v_1$')
axes[1,0].plot(syst.T, syst.get_values('P_t'), label=f'$P_t$')
axes[1,1].plot(syst.T, syst.get_values('Q_t'), label=f'$Q_t$')
ssa.eval_ss(syst)
from scipy.signal import ss2tf,lti,bode
num,den =ss2tf(syst.A,syst.B,syst.C,syst.D,input=0)
G = lti(num[1],den)
w, mag, phase = G.bode()
plt.figure()
plt.semilogx(w, mag) # Bode magnitude plot
plt.figure()
plt.semilogx(w, phase) # Bode phase plot
plt.show()
events=[{'t_end':1.0,'P_t':0.8, 'Q_t':0.5},
{'t_end':10.0, 'p_m':0.9}
]
syst.simulate(events,xy0=1.0);
syst.inputs_run_list
0.01/6
syst.B
syst.struct[0]['Fu']
###Output
_____no_output_____ |
utils/notebooks/model_full.ipynb | ###Markdown
Shuffle the dataset.
###Code
from sklearn.utils import shuffle
processed_data = shuffle(processed_data)
X = processed_data[:,:-1].astype(float)
y = processed_data[:,-1].astype(int)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.1)
Xtrain[0].shape[0]
print(Xtrain.dtype)
print(Ytrain)
###Output
[0 0 0 ... 0 0 0]
###Markdown
Fully connected neural network model with 3 hidden layers:
###Code
import tensorflow as tf
model = tf.keras.models.Sequential([
tf.keras.Input(shape=(X_train[0].shape[0],)), #Xtrain[0].shape[0] = 108 -> input size
tf.keras.layers.Dense(15, activation='relu'), #represents 1st hidden layer
tf.keras.layers.Dropout(0.2), # dropout regularization with dropout probability 20% percent
tf.keras.layers.Dense(7, activation='relu'), #represents the 2nd hidden layer
tf.keras.layers.Dropout(0.2), # dropout regularization with dropout probability 20% percent
tf.keras.layers.Dense(4, activation='relu'), #represents the 3rd hidden layer
tf.keras.layers.Dropout(0.2), # dropout regularization with dropout probability 20% percent
tf.keras.layers.Dense(1, activation='sigmoid') # sigmoid activation at output since it is binary classification
])
model.compile(optimizer='adam',
loss='binary_crossentropy',
metrics=['accuracy'])
model.summary()
###Output
Model: "sequential_7"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_28 (Dense) (None, 15) 1635
_________________________________________________________________
dropout_21 (Dropout) (None, 15) 0
_________________________________________________________________
dense_29 (Dense) (None, 7) 112
_________________________________________________________________
dropout_22 (Dropout) (None, 7) 0
_________________________________________________________________
dense_30 (Dense) (None, 4) 32
_________________________________________________________________
dropout_23 (Dropout) (None, 4) 0
_________________________________________________________________
dense_31 (Dense) (None, 1) 5
=================================================================
Total params: 1,784
Trainable params: 1,784
Non-trainable params: 0
_________________________________________________________________
###Markdown
Train the model with 10 epochs:
###Code
r2 = model.fit(X_train, y_train, epochs=150)
###Output
Epoch 1/150
101/101 [==============================] - 1s 1ms/step - loss: 145.9402 - accuracy: 0.8358
Epoch 2/150
101/101 [==============================] - 0s 1ms/step - loss: 30.3910 - accuracy: 0.7783
Epoch 3/150
101/101 [==============================] - 0s 960us/step - loss: 20.8436 - accuracy: 0.8890
Epoch 4/150
101/101 [==============================] - 0s 960us/step - loss: 12.8123 - accuracy: 0.8991
Epoch 5/150
101/101 [==============================] - 0s 990us/step - loss: 11.4324 - accuracy: 0.9137
Epoch 6/150
101/101 [==============================] - 0s 1ms/step - loss: 4.3560 - accuracy: 0.9117
Epoch 7/150
101/101 [==============================] - 0s 1000us/step - loss: 2.3523 - accuracy: 0.9109
Epoch 8/150
101/101 [==============================] - 0s 1ms/step - loss: 3.8448 - accuracy: 0.9153
Epoch 9/150
101/101 [==============================] - 0s 1ms/step - loss: 1.5151 - accuracy: 0.9146
Epoch 10/150
101/101 [==============================] - 0s 1ms/step - loss: 1.0658 - accuracy: 0.9061
Epoch 11/150
101/101 [==============================] - 0s 1ms/step - loss: 0.6152 - accuracy: 0.9204: 0s - loss: 0.6139 - accuracy: 0.92
Epoch 12/150
101/101 [==============================] - 0s 1ms/step - loss: 0.7392 - accuracy: 0.9057
Epoch 13/150
101/101 [==============================] - 0s 1ms/step - loss: 1.3826 - accuracy: 0.9137
Epoch 14/150
101/101 [==============================] - 0s 1ms/step - loss: 0.5601 - accuracy: 0.9182
Epoch 15/150
101/101 [==============================] - 0s 1ms/step - loss: 0.4373 - accuracy: 0.9223
Epoch 16/150
101/101 [==============================] - 0s 950us/step - loss: 0.7177 - accuracy: 0.9190
Epoch 17/150
101/101 [==============================] - 0s 940us/step - loss: 0.3467 - accuracy: 0.9114
Epoch 18/150
101/101 [==============================] - 0s 940us/step - loss: 0.3273 - accuracy: 0.9154
Epoch 19/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3264 - accuracy: 0.9112
Epoch 20/150
101/101 [==============================] - 0s 920us/step - loss: 0.3304 - accuracy: 0.9062
Epoch 21/150
101/101 [==============================] - 0s 900us/step - loss: 0.3189 - accuracy: 0.9109
Epoch 22/150
101/101 [==============================] - 0s 890us/step - loss: 0.3067 - accuracy: 0.9160
Epoch 23/150
101/101 [==============================] - 0s 910us/step - loss: 0.3122 - accuracy: 0.9112
Epoch 24/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3101 - accuracy: 0.9113
Epoch 25/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2955 - accuracy: 0.9180
Epoch 26/150
101/101 [==============================] - 0s 960us/step - loss: 0.2911 - accuracy: 0.9191
Epoch 27/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3017 - accuracy: 0.9128
Epoch 28/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2881 - accuracy: 0.9190
Epoch 29/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3068 - accuracy: 0.9093
Epoch 30/150
101/101 [==============================] - 0s 910us/step - loss: 0.2924 - accuracy: 0.9161
Epoch 31/150
101/101 [==============================] - 0s 910us/step - loss: 0.2855 - accuracy: 0.9190
Epoch 32/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3017 - accuracy: 0.9110
Epoch 33/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2944 - accuracy: 0.9142
Epoch 34/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3004 - accuracy: 0.9112
Epoch 35/150
101/101 [==============================] - 0s 960us/step - loss: 0.2884 - accuracy: 0.9165
Epoch 36/150
101/101 [==============================] - 0s 990us/step - loss: 0.2977 - accuracy: 0.9123
Epoch 37/150
101/101 [==============================] - 0s 910us/step - loss: 0.2839 - accuracy: 0.9183
Epoch 38/150
101/101 [==============================] - 0s 910us/step - loss: 0.3008 - accuracy: 0.9105
Epoch 39/150
101/101 [==============================] - 0s 830us/step - loss: 0.2925 - accuracy: 0.9144
Epoch 40/150
101/101 [==============================] - 0s 830us/step - loss: 0.2934 - accuracy: 0.9140
Epoch 41/150
101/101 [==============================] - 0s 820us/step - loss: 0.2894 - accuracy: 0.9157
Epoch 42/150
101/101 [==============================] - 0s 850us/step - loss: 0.2849 - accuracy: 0.9174
Epoch 43/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2896 - accuracy: 0.9154
Epoch 44/150
101/101 [==============================] - 0s 950us/step - loss: 0.2937 - accuracy: 0.9136
Epoch 45/150
101/101 [==============================] - 0s 940us/step - loss: 0.3066 - accuracy: 0.9080
Epoch 46/150
101/101 [==============================] - 0s 890us/step - loss: 0.3024 - accuracy: 0.9099
Epoch 47/150
101/101 [==============================] - 0s 900us/step - loss: 0.2744 - accuracy: 0.9218
Epoch 48/150
101/101 [==============================] - 0s 917us/step - loss: 0.2914 - accuracy: 0.9146
Epoch 49/150
101/101 [==============================] - 0s 842us/step - loss: 0.2777 - accuracy: 0.9204
Epoch 50/150
101/101 [==============================] - 0s 884us/step - loss: 0.3010 - accuracy: 0.9104
Epoch 51/150
101/101 [==============================] - 0s 844us/step - loss: 0.2773 - accuracy: 0.9205
Epoch 52/150
101/101 [==============================] - 0s 906us/step - loss: 0.2917 - accuracy: 0.9144
Epoch 53/150
101/101 [==============================] - 0s 854us/step - loss: 0.2798 - accuracy: 0.9196
Epoch 54/150
101/101 [==============================] - 0s 888us/step - loss: 0.2918 - accuracy: 0.9144
Epoch 55/150
101/101 [==============================] - 0s 940us/step - loss: 0.2983 - accuracy: 0.9116
Epoch 56/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2943 - accuracy: 0.9135
Epoch 57/150
101/101 [==============================] - 0s 980us/step - loss: 0.3071 - accuracy: 0.9079
Epoch 58/150
101/101 [==============================] - 0s 879us/step - loss: 0.2840 - accuracy: 0.9178
Epoch 59/150
101/101 [==============================] - 0s 860us/step - loss: 0.3051 - accuracy: 0.9087
Epoch 60/150
101/101 [==============================] - 0s 842us/step - loss: 0.2954 - accuracy: 0.9129
Epoch 61/150
101/101 [==============================] - 0s 845us/step - loss: 0.2985 - accuracy: 0.9115
Epoch 62/150
101/101 [==============================] - 0s 853us/step - loss: 0.2796 - accuracy: 0.9194
Epoch 63/150
101/101 [==============================] - 0s 833us/step - loss: 0.2559 - accuracy: 0.9297
Epoch 64/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2882 - accuracy: 0.9159
Epoch 65/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3027 - accuracy: 0.9097
Epoch 66/150
101/101 [==============================] - 0s 880us/step - loss: 0.2836 - accuracy: 0.9179
Epoch 67/150
101/101 [==============================] - 0s 840us/step - loss: 0.2836 - accuracy: 0.9179
Epoch 68/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2792 - accuracy: 0.9198
Epoch 69/150
101/101 [==============================] - 0s 1ms/step - loss: 0.3025 - accuracy: 0.9099
Epoch 70/150
101/101 [==============================] - 0s 850us/step - loss: 0.3135 - accuracy: 0.9052
Epoch 71/150
101/101 [==============================] - 0s 863us/step - loss: 0.2897 - accuracy: 0.9153
Epoch 72/150
101/101 [==============================] - 0s 839us/step - loss: 0.2944 - accuracy: 0.9133
Epoch 73/150
101/101 [==============================] - 0s 894us/step - loss: 0.2931 - accuracy: 0.9139
Epoch 74/150
101/101 [==============================] - 0s 852us/step - loss: 0.3052 - accuracy: 0.9087
Epoch 75/150
101/101 [==============================] - 0s 900us/step - loss: 0.2947 - accuracy: 0.9132
Epoch 76/150
101/101 [==============================] - 0s 890us/step - loss: 0.2757 - accuracy: 0.9212
Epoch 77/150
101/101 [==============================] - 0s 880us/step - loss: 0.2985 - accuracy: 0.9116
Epoch 78/150
101/101 [==============================] - 0s 1ms/step - loss: 0.2873 - accuracy: 0.9163
Epoch 79/150
101/101 [==============================] - 0s 910us/step - loss: 0.3000 - accuracy: 0.9109
###Markdown
Plot the losses:
###Code
plt.plot(r2.history['loss'], label='loss')
#plt.plot(r2.history['val_loss'], label='val_loss')
plt.legend()
###Output
_____no_output_____
###Markdown
Plot the accuracies:
###Code
plt.plot(r2.history['accuracy'], label='acc')
#plt.plot(r2.history['val_accuracy'], label='val_acc')
plt.legend()
print(model.evaluate(X_test, y_test))
model.save_weights('./weights/model_systolic_diastolic_excluded/')
###Output
_____no_output_____
###Markdown
**Applying this dataset on different models to create benchmarks to compare with Fully Connected model:** *Logistic Regression* **(linear classifier):**
###Code
from sklearn.linear_model import LogisticRegression
model =LogisticRegression()
model.fit(X_train, y_train)
print(model.score(X_train, y_train))
print(model.score(X_test,y_test))
for i in range(X_train[0].shape[0]):
print("weight for the feature",i,":",model.coef_[0][i])
model.coef_[0].sort()
print("sorted list, positive weights contribute MORE TO DETERMINE potential hypertension case and negative features conribute more to make prediction as NOT hypertension. So we can say 0 valued weights do not contribute to anything:",model.coef_[0])
###Output
weight for the feature 0 : -0.002204197745180158
weight for the feature 1 : -0.0020580556545120496
weight for the feature 2 : -0.0018476029920989035
weight for the feature 3 : -0.0017392015573981473
weight for the feature 4 : -0.001309569884951161
weight for the feature 5 : -0.0012986945766602534
weight for the feature 6 : -0.0012558997868359195
weight for the feature 7 : -0.001044450183718869
weight for the feature 8 : -0.0010411694398086097
weight for the feature 9 : -0.001003031407691053
weight for the feature 10 : -0.0009978758145447076
weight for the feature 11 : -0.0008635668161326685
weight for the feature 12 : -0.0008060209811952334
weight for the feature 13 : -0.0007723014259508736
weight for the feature 14 : -0.00043181865850317993
weight for the feature 15 : -0.00036305971840124495
weight for the feature 16 : -0.00034605420873229067
weight for the feature 17 : -0.0003215899842853878
weight for the feature 18 : -0.00029768026994142743
weight for the feature 19 : -0.00021503252902479187
weight for the feature 20 : -0.00018414498287057135
weight for the feature 21 : -0.00015214187901898255
weight for the feature 22 : -0.00013145456687364403
weight for the feature 23 : -9.131808199608684e-05
weight for the feature 24 : -9.069291103926271e-05
weight for the feature 25 : -6.26050637143378e-05
weight for the feature 26 : -5.234520575306838e-05
weight for the feature 27 : -4.733510538892756e-05
weight for the feature 28 : -2.7441116057177076e-05
weight for the feature 29 : -2.112005390565407e-05
weight for the feature 30 : -1.9644759250143093e-05
weight for the feature 31 : -1.6648835251829046e-05
weight for the feature 32 : -1.438450096272556e-05
weight for the feature 33 : -1.4192585289277e-05
weight for the feature 34 : -1.4041359495178202e-05
weight for the feature 35 : -1.2957218149275538e-05
weight for the feature 36 : -1.2701743479999902e-05
weight for the feature 37 : -1.2260372977818067e-05
weight for the feature 38 : -1.2089988823209605e-05
weight for the feature 39 : -1.1456858760496845e-05
weight for the feature 40 : -1.1426565738571777e-05
weight for the feature 41 : -1.1100513127682143e-05
weight for the feature 42 : -1.1045382641201871e-05
weight for the feature 43 : -1.0900586338660222e-05
weight for the feature 44 : -1.0646041057854768e-05
weight for the feature 45 : -8.895829558217836e-06
weight for the feature 46 : -8.467076640820137e-06
weight for the feature 47 : -8.46683230710532e-06
weight for the feature 48 : -7.464918992053486e-06
weight for the feature 49 : -6.660404497702645e-06
weight for the feature 50 : -6.660404497702637e-06
weight for the feature 51 : -5.585068607032158e-06
weight for the feature 52 : -5.585068607032158e-06
weight for the feature 53 : -5.085869355470354e-06
weight for the feature 54 : -3.940612734408723e-06
weight for the feature 55 : -3.684987853060102e-06
weight for the feature 56 : -3.5138101710966967e-06
weight for the feature 57 : -3.433581989436847e-06
weight for the feature 58 : -2.8597817095667753e-06
weight for the feature 59 : -2.3503274583278408e-06
weight for the feature 60 : -2.3302029746963264e-06
weight for the feature 61 : -1.9198180225464525e-06
weight for the feature 62 : -1.7098052967395282e-06
weight for the feature 63 : -1.1475896612664792e-06
weight for the feature 64 : -1.127629282336915e-06
weight for the feature 65 : -1.0988664126315144e-06
weight for the feature 66 : -9.187410055737733e-07
weight for the feature 67 : -8.706546627976764e-07
weight for the feature 68 : -7.266318166202316e-07
weight for the feature 69 : -5.991694140895009e-07
weight for the feature 70 : -5.591708630871402e-07
weight for the feature 71 : -5.270349995785656e-07
weight for the feature 72 : -5.270349995785654e-07
weight for the feature 73 : -5.127387071059556e-07
weight for the feature 74 : -4.1192352710556633e-07
weight for the feature 75 : -3.4481447533078734e-07
weight for the feature 76 : -3.194320807046258e-07
weight for the feature 77 : -2.9422528169894045e-07
weight for the feature 78 : -2.2621897169226073e-07
weight for the feature 79 : -1.9729710334979552e-07
weight for the feature 80 : -9.22755470703345e-08
weight for the feature 81 : 4.9322132183351365e-08
weight for the feature 82 : 2.269975263746754e-07
weight for the feature 83 : 2.955866648387692e-07
weight for the feature 84 : 4.3215383668249797e-07
weight for the feature 85 : 4.3215383668249797e-07
weight for the feature 86 : 4.5196322290104333e-07
weight for the feature 87 : 4.562703771415081e-07
weight for the feature 88 : 6.443546015296859e-07
weight for the feature 89 : 6.502938593458788e-07
weight for the feature 90 : 7.117450453589198e-07
weight for the feature 91 : 8.54650329545992e-07
weight for the feature 92 : 9.64787197472597e-07
weight for the feature 93 : 1.0149093497867197e-06
weight for the feature 94 : 1.0262422297789211e-06
weight for the feature 95 : 1.1670970349930403e-06
weight for the feature 96 : 1.3989427603701783e-06
weight for the feature 97 : 1.6152415296342415e-06
weight for the feature 98 : 1.6152415296342415e-06
weight for the feature 99 : 1.7844959395724763e-06
weight for the feature 100 : 1.9143781245295206e-06
weight for the feature 101 : 2.1390278579975307e-06
weight for the feature 102 : 2.195553946908131e-06
weight for the feature 103 : 2.9235156724174116e-06
weight for the feature 104 : 3.484529034590651e-06
weight for the feature 105 : 4.617190687997532e-06
weight for the feature 106 : 7.909793602692546e-06
weight for the feature 107 : 9.036461893761524e-06
sorted list, positive weights contribute MORE TO DETERMINE potential fraud case and negative features conribute to make prediction as NOT FRAUD. So we can say 0 valued weights do not contribute to anything: [-2.20419775e-03 -2.05805565e-03 -1.84760299e-03 -1.73920156e-03
-1.30956988e-03 -1.29869458e-03 -1.25589979e-03 -1.04445018e-03
-1.04116944e-03 -1.00303141e-03 -9.97875815e-04 -8.63566816e-04
-8.06020981e-04 -7.72301426e-04 -4.31818659e-04 -3.63059718e-04
-3.46054209e-04 -3.21589984e-04 -2.97680270e-04 -2.15032529e-04
-1.84144983e-04 -1.52141879e-04 -1.31454567e-04 -9.13180820e-05
-9.06929110e-05 -6.26050637e-05 -5.23452058e-05 -4.73351054e-05
-2.74411161e-05 -2.11200539e-05 -1.96447593e-05 -1.66488353e-05
-1.43845010e-05 -1.41925853e-05 -1.40413595e-05 -1.29572181e-05
-1.27017435e-05 -1.22603730e-05 -1.20899888e-05 -1.14568588e-05
-1.14265657e-05 -1.11005131e-05 -1.10453826e-05 -1.09005863e-05
-1.06460411e-05 -8.89582956e-06 -8.46707664e-06 -8.46683231e-06
-7.46491899e-06 -6.66040450e-06 -6.66040450e-06 -5.58506861e-06
-5.58506861e-06 -5.08586936e-06 -3.94061273e-06 -3.68498785e-06
-3.51381017e-06 -3.43358199e-06 -2.85978171e-06 -2.35032746e-06
-2.33020297e-06 -1.91981802e-06 -1.70980530e-06 -1.14758966e-06
-1.12762928e-06 -1.09886641e-06 -9.18741006e-07 -8.70654663e-07
-7.26631817e-07 -5.99169414e-07 -5.59170863e-07 -5.27035000e-07
-5.27035000e-07 -5.12738707e-07 -4.11923527e-07 -3.44814475e-07
-3.19432081e-07 -2.94225282e-07 -2.26218972e-07 -1.97297103e-07
-9.22755471e-08 4.93221322e-08 2.26997526e-07 2.95586665e-07
4.32153837e-07 4.32153837e-07 4.51963223e-07 4.56270377e-07
6.44354602e-07 6.50293859e-07 7.11745045e-07 8.54650330e-07
9.64787197e-07 1.01490935e-06 1.02624223e-06 1.16709703e-06
1.39894276e-06 1.61524153e-06 1.61524153e-06 1.78449594e-06
1.91437812e-06 2.13902786e-06 2.19555395e-06 2.92351567e-06
3.48452903e-06 4.61719069e-06 7.90979360e-06 9.03646189e-06]
###Markdown
*Decision Tree*
###Code
from sklearn.tree import DecisionTreeClassifier
model =DecisionTreeClassifier()
model.fit(X_train, y_train)
print(model.score(X_train, y_train))
print(model.score(X_test,y_test))
###Output
1.0
1.0
###Markdown
*AdaBoost*
###Code
from sklearn.ensemble import AdaBoostClassifier
model =AdaBoostClassifier()
model.fit(X_train, y_train)
print(model.score(X_train, y_train))
print(model.score(X_test,y_test))
# Visualize the data (tsne is great but slow.)
from sklearn.manifold import TSNE
tsne = TSNE(n_components=2)
transformed = tsne.fit_transform(X_train)
#visualize in the 2d
plt.scatter(transformed[:,0], transformed[:,1], c=y_train) #take first two columns in order to get 2d plot.
plt.show()
transformed = tsne.fit_transform(X_test)
# visualize the clouds in 2-D
plt.scatter(transformed[:,0], transformed[:,1], c=y_test) #take first two columns in order to get 2d plot.
plt.show()
###Output
_____no_output_____ |
temp/HowToBreakIntoTheField.ipynb | ###Markdown
ScreencastIn the previous video, I brought a few questions we will be exploring throughout this lesson. First, let's take a look at the data, and see how we might answer the first question about how to break into the field of becoming a software developoer according to the survey results.To get started, let's read in the necessary libraries we will need to wrangle our data: pandas and numpy. If we decided to build some basic plots, matplotlib might prove useful as well.
###Code
import numpy as np
import pandas as pd
from collections import defaultdict
import matplotlib.pyplot as plt
%matplotlib inline
df = pd.read_csv('./Part I/stackoverflow/survey_results_public.csv')
df.head()
###Output
_____no_output_____
###Markdown
Now to look at our first question of interest: What do those employed in the industry suggest to help others enter the field? Looking at the `CousinEducation` field, you can see what these individuals would suggest to help others break into their field. Below you can take a look at the full field that survey participants would see.
###Code
df2 = pd.read_csv('./Part I/stackoverflow/survey_results_schema.csv')
list(df2[df2.Column == 'CousinEducation']['Question'])
#Let's have a look at what the participants say
study = df['CousinEducation'].value_counts().reset_index()
study.head()
# Oh this isn't what I was expecting, it is grouping items together if a participant provided
# more than just one answer. Let's see if we can clean this up.
# first to change this index column to a more appropriate name
study.rename(columns={'index': 'method', 'CousinEducation': 'count'}, inplace=True)
study.head()
###Output
_____no_output_____
###Markdown
A quick look through data, allows us to create a list of all of the individual methods marked by a user.
###Code
# Here is a list of the different answers provided
possible_vals = ["Take online courses", "Buy books and work through the exercises",
"None of these", "Part-time/evening courses", "Return to college",
"Contribute to open source", "Conferences/meet-ups", "Bootcamp",
"Get a job as a QA tester", "Participate in online coding competitions",
"Master's degree", "Participate in hackathons", "Other"]
#Now we want to see how often each of these individual values appears - I wrote
# this function to assist with process - it isn't the best solution, but it gets
# the job done and our dataset isn't large enough to computationally hurt us too much.
def total_count(df, col1, col2, look_for):
'''
INPUT:
df - the pandas dataframe you want to search
col1 - the column name you want to look through
col2 - the column you want to count values from
look_for - a list of strings you want to search for in each row of df[col]
OUTPUT:
new_df - a dataframe of each look_for with the count of how often it shows up
'''
new_df = defaultdict(int)
for val in look_for:
for idx in range(df.shape[0]):
if val in df[col1][idx]:
new_df[val] += int(df[col2][idx])
new_df = pd.DataFrame(pd.Series(new_df)).reset_index()
new_df.columns = [col1, col2]
new_df.sort_values('count', ascending=False, inplace=True)
return new_df
# Now we can use our function and take a look at the results
# Looks like good news for Udacity - most individuals think that you
# should take online courses
study_df = total_count(study, 'method', 'count', possible_vals)
study_df
# We might also look at the percent
study_df['perc'] = study_df['count']/np.sum(study_df['count'])
study_df
###Output
_____no_output_____
###Markdown
We might want to take this one step further and say we care more about the methods that are suggested by those who earn more, or those who have higher job satisfaction. Let's take a stab at incorporating that into this analysis.
###Code
# let's rewrite part of this function to get the mean salary for each method
def mean_amt(df, col_name, col_mean, look_for):
'''
INPUT:
df - the pandas dataframe you want to search
col_name - the column name you want to look through
col_count - the column you want to count values from
col_mean - the column you want the mean amount for
look_for - a list of strings you want to search for in each row of df[col]
OUTPUT:
df_all - holds sum, square, total, mean, variance, and standard deviation for the col_mean
'''
new_df = defaultdict(int)
squares_df = defaultdict(int)
denoms = dict()
for val in look_for:
denoms[val] = 0
for idx in range(df.shape[0]):
if df[col_name].isnull()[idx] == False:
if val in df[col_name][idx] and df[col_mean][idx] > 0:
new_df[val] += df[col_mean][idx]
squares_df[val] += df[col_mean][idx]**2 #Needed to understand the spread
denoms[val] += 1
# Turn into dataframes
new_df = pd.DataFrame(pd.Series(new_df)).reset_index()
squares_df = pd.DataFrame(pd.Series(squares_df)).reset_index()
denoms = pd.DataFrame(pd.Series(denoms)).reset_index()
# Change the column names
new_df.columns = [col_name, 'col_sum']
squares_df.columns = [col_name, 'col_squares']
denoms.columns = [col_name, 'col_total']
# Merge dataframes
df_means = pd.merge(new_df, denoms)
df_all = pd.merge(df_means, squares_df)
# Additional columns needed for analysis
df_all['mean_col'] = df_means['col_sum']/df_means['col_total']
df_all['var_col'] = df_all['col_squares']/df_all['col_total'] - df_all['mean_col']**2
df_all['std_col'] = np.sqrt(df_all['var_col'])
df_all['lower_95'] = df_all['mean_col'] - 1.96*df_all['std_col']/np.sqrt(df_all['col_total'])
df_all['upper_95'] = df_all['mean_col'] + 1.96*df_all['std_col']/np.sqrt(df_all['col_total'])
return df_all
df_all = mean_amt(df, 'CousinEducation', 'Salary', possible_vals)
# To get a simple answer to our questions - see these two tables.
df_all.sort_values('mean_col', ascending=False)
study_df
###Output
_____no_output_____
###Markdown
Although we can see the mean salary is highest for the individuals who say that you should contribute to open source, you might be asking - is that really a significant difference? The salary differences don't see that large...By the Central Limit Theorem, we know that the mean of any set of data will follow a normal distribution with a standard deviation equal to the standard deviation of the original data divided by the square root of the sample size, as long as we collect a large enough sample size. With that in mind, we can consider two salaries significantly different if a second salary is two standard deviations or more away from the other.Using the lower and upper bound components, we can get an idea of the salaries that are significantly different from one another.
###Code
# Quiz - perform a similar analysis looking at career and job satisfaction for this individuals
# to determine which you want to be like
df_jobsat = mean_amt(df, 'CousinEducation', 'JobSatisfaction', possible_vals)
df_jobsat.sort_values('mean_col', ascending=False)
pd.DataFrame(np.hstack([df_jobsat, df_all]))
pd.DataFrame?
df_jobsat.col_total
df_dotplot = pd.DataFrame(np.array(['Method', "At least Master's", "Less Than Master's",
"Master's Degree",0.0589517,0.0293459,
"Bootcamp",0.0746172,0.071824,
"Become QA Tester",0.0484688,0.0457388,
"Buy Books",0.162073,0.161205,
"None of these",0.00836278,0.00827705,
"Part Time Courses",0.103298,0.103248,
"Return to College",0.0687279,0.0689754,
"Online Courses",0.207892,0.2099,
"Contribute to Opensource",0.097821,0.10323,
"Coding Competitions",0.0453475,0.0508806,
"Other",0.0269729,0.0338607,
"Hackathons", 0.0316254,0.0395937,
"Conferences", 0.0658422, 0.0739201]).reshape((14, 3)))
df_dotplot.columns = df_dotplot.iloc[0]
df_dotplot.drop(0, inplace=True)
import seaborn as sns
df_dotplot.prop = df_dotplot.prop.astype(float)
df_dotplot = df_dotplot.melt(id_vars='Method', value_name='prop', var_name='status')
sns.pointplot(data=df_dotplot, x='status', y='prop', hue='Method');
for idx, row in df_dotplot.iterrows():
plt.plot([0,1],[row["At least Master's"], row["Less Than Master's"]]);
plt.text(1.05, row["Less Than Master's"], row['Method']);
plt.xticks([0,1], ["At least Master's", "Less Than Master's"]);
for row in df_dotplot:
print(row)
###Output
Method
At least Master's
Less Than Master's
|
SpeedDatingAnalysis/Project_Assignment.ipynb | ###Markdown
Speed Dating Data Analysis Team Members: Onyinye Ihedero, Sai Manoj Gaddipati, Esha Somavarapu Source of Data: www.kaggle.com Packages Used : Numpy, Pandas, Tabpy(not full fledged) Softwares Used: Anaconda, Jupyter Notebook in Anaconda, Tableau Software, MS-Office.
###Code
import numpy as np
import pandas as pd
data=pd.read_table("speedData.csv", sep=',')
data_reqRow=data[['age', 'gender', 'field_cd', 'race', 'imprace', 'imprelig', 'income', 'goal','date','samerace' , 'race_o', 'go_out', 'career_c', 'attr1_1', 'sinc1_1', 'intel1_1',
'fun1_1', 'amb1_1', 'shar1_1', 'satis_2', 'attr7_2', 'sinc7_2', 'intel7_2', 'fun7_2', 'amb7_2', 'shar7_2', 'attr1_s', 'sinc1_s', 'intel1_s', 'fun1_s', 'amb1_s', 'shar1_s', 'you_call', 'them_cal', 'date_3',
'num_in_3', 'attr7_3', 'sinc7_3', 'intel7_3', 'fun7_3', 'amb7_3', 'shar7_3']]
import numpy as np
data_reqRow = data_reqRow[np.logical_not(np.isnan(data_reqRow.attr1_s))]
data_reqRow = data_reqRow.reset_index(drop=True)
data_reqRow
#To output to .csv do this
data_reqRow.to_csv('OutputFinal.csv', index=False)
#doing group by ages and getting the count for each one
data_count_age=data_reqRow.groupby('age').agg('count').reset_index()
data_count_age#.head()
#notes for the output below:
#we can see that the attendance is more for ages between 22 and 30. The population for ages beyond that is sparse.
#doing group by gender and getting the count for each one
data_count_gender=data_reqRow.groupby('gender').agg('count').reset_index()
data_count_gender#.head()
#notes:
#number of females=4141
#number of males=4174
#doing group by race and getting the count for each one
data_count_race=data_reqRow.groupby('race').agg('count').reset_index()
data_count_race#.head()
#notes on results:
#heres the race description:
#Black/African American=1 (420)
#European/Caucasian-American=2 (4727)
#Latino/Hispanic American=3 (664)
#Asian/Pacific Islander/Asian-American=4 (1982)
#Native American=5 (0)
#Other=6 (522)
#it can be noted that the population of Native Americans is 0, no one attended.
#doing group by categorised careers and getting the count for each one
data_count_career=data_reqRow.groupby('career_c').agg('count').reset_index()
data_count_career#.head()
#the count results are written in brackets
#classification details for career:
#1= Lawyer
#2= Academic/Research
#3= Psychologist
#4= Doctor/Medicine
#5=Engineer
#6= Creative Arts/Entertainment
#7= Banking/Consulting/Finance/Marketing/
# Business/CEO/Entrepreneur/Admin
#8= Real Estate
#9= International/Humanitarian Affairs
#10= Undecided
#11=Social Work
#12=Speech Pathology
#13=Politics
#14=Pro sports/Athletics
#15=Other
#16=Journalism
#17=Architecture
###Output
_____no_output_____
###Markdown
I am adding samerace and race_o back into the dataset at the beginning itself.
###Code
#doing group by "if preferring partner from same race" and getting the count for each one
data_count_samerace=data_reqRow.groupby('samerace').agg('count').reset_index()
data_count_samerace#.head()
#notes:
#
#doing group by categorised careers and getting the count for each one
data_count_partner_race=data_reqRow.groupby('race_o').agg('count').reset_index()
data_count_partner_race#.head()
#which race thinks it is more important to date the same race:
res=data_reqRow.pivot_table(index='race', values='imprace', aggfunc=np.mean)
print(res)
res.to_csv('raceVimprace.csv')
#Note on Results:
#no race thinks it is extremely important to date someone form the same race.
#the only race that feels more important to date their same race is:
#European-Caucasian/American with an average imprace of 4.29
#which age thinks it is more important to date the same race:
res2=data_reqRow.pivot_table(index='age', values='imprace', aggfunc=np.mean)
res2.to_csv('ageVimprace.csv')
res2
#there is a strong response from ages 35, 37, and 42
#although the number of people who participated is less for these age groups
#which age thinks it is more important to date the same race:
res3=data_reqRow.pivot_table(index='race', values='imprelig', aggfunc=np.mean)
res3.to_csv('ageVimprace.csv')
res3
#which age thinks it is more important to date people from same religion:
res4=data_reqRow.pivot_table(index='age', values='imprelig', aggfunc=np.mean)
res4.to_csv('ageVimprelig.csv')
res4
res6=data_reqRow.pivot_table(index='gender', values=['attr1_1', 'sinc1_1', 'intel1_1', 'fun1_1', 'amb1_1', 'shar1_1'], aggfunc=np.mean)
res5.to_csv('start.csv')
res6
res5=data_reqRow.pivot_table(index='gender', values=['attr7_2', 'sinc7_2', 'intel7_2', 'fun7_2', 'amb7_2', 'shar7_2'], aggfunc=np.mean)
res5.to_csv('after.csv')
res5
#we can do the same thing for which race prefers which race but I am a little confused about which variable is the preferred race varaiable4
# here are the variables I am confused about:
#samerace: participant and the partner were the same race. 1= yes, 0=no
#age_o: age of partner
#race_o: race of partner
#pf_o_att: partner’s stated preference at Time 1 (attr1_1) for all 6 attributes
data_reqRow.to_csv('data_with.csv')
###Output
_____no_output_____ |
notebooks/PhysionetCharacterization/Explore Physionet CTG Outcome Metadata Apgar 1.ipynb | ###Markdown
Explore Physionet CTG Outcome Metadata Apgar 1see: https://physionet.org/physiobank/database/ctu-uhb-ctgdb/Includes:- Plots of various outcome metrics vs apgar 1- Regression models to predict apgar1 from other outcome metrics
###Code
import config_local
from config_common import *
import wfdb
import os
from pprint import pprint
import numpy as np
import matplotlib.pyplot as plt
import scipy
import scipy.signal
from sklearn import svm
from sklearn.decomposition import PCA
import math
import collections
from ctg_utils import get_all_recno, parse_meta_comments
import random
###Output
_____no_output_____
###Markdown
Config Code
###Code
def jitter(x, w=2):
return x + (random.random()-0.5)/w
def display_metric_vs_apgar(metric, apgar, title='', xlabel='', ylabel='',
limits=None, thresh=None):
plt.figure(figsize=(5,5))
if title:
plt.title(title)
plt.scatter(metric, [jitter(x) for x in apgar], s=3)
if thresh:
plt.plot([thresh, thresh], [0, 10], 'r--')
plt.ylim(0, 10)
if limits:
plt.plot(limits, [7, 7], 'r--')
plt.xlim(*limits)
if ylabel:
plt.ylabel(ylabel)
if apgar:
plt.xlabel(xlabel)
plt.show()
def display_predictions(actual, pred, title=None, xlabel='Apgar 5'):
all_error = [abs(val-pred[i]) for i, val in enumerate(actual)]
print('Error -- mean: {:0.2f} std: {:0.2f}'.format(np.mean(all_error), np.std(all_error)))
plt.figure(figsize=(5,5))
if title:
plt.title(title)
plt.scatter([jitter(x) for x in actual], [x for x in pred], s=3)
plt.ylim(0, 10)
plt.xlim(0, 10)
plt.plot([0,10], [0, 10], 'g--', alpha=0.25)
plt.plot([0,9.5], [0.5, 10], 'r--', alpha=0.25)
plt.plot([0.5,10], [0, 9.5], 'r--', alpha=0.25)
plt.plot([7,7], [0, 10], 'r--', alpha=0.25)
plt.plot([0, 10], [7,7], 'r--', alpha=0.25)
plt.ylabel('Predicted {}'.format(xlabel))
plt.xlabel(xlabel)
plt.show()
###Output
_____no_output_____
###Markdown
Gather All Recording Metadata
###Code
all_meta = {}
all_error = []
for recno in sorted(get_all_recno(media_recordings_dir_full)):
recno_full = os.path.join(media_recordings_dir_full, recno)
#print('Record: {}'.format(recno))
try:
all_sig, meta = wfdb.io.rdsamp(recno_full)
meta['comments'] = parse_meta_comments(meta['comments'])
all_meta[recno] = meta['comments']
except Exception as e:
print(' Error: {}'.format(e))
all_error.append(recno)
###Output
_____no_output_____
###Markdown
Filter nan values
###Code
for recno in sorted(all_meta.keys()):
entry = all_meta[recno]['Outcome']
if math.isnan(entry['BDecf']):
print('{}: Recording contains NaN'.format(recno))
del all_meta[recno]
###Output
1044: Recording contains NaN
1070: Recording contains NaN
1211: Recording contains NaN
1215: Recording contains NaN
1356: Recording contains NaN
1373: Recording contains NaN
1383: Recording contains NaN
1419: Recording contains NaN
2006: Recording contains NaN
2034: Recording contains NaN
2046: Recording contains NaN
###Markdown
Outcomes
###Code
for recno in sorted(all_meta.keys()):
# recno = '1001'
entry = all_meta[recno]['Outcome']
print('{:4}: pH:{:5.2f} BDecf: {:5.2f} pCO2: {:5.2f} BE: {:6.2f} Apgar1: {:2} Apgar5: {:2}'.format(
recno, entry['pH'], entry['BDecf'], entry['pCO2'], entry['BE'], entry['Apgar1'], entry['Apgar5']))
all_pH = [entry['Outcome']['pH'] for entry in all_meta.values()]
all_BDecf = [entry['Outcome']['BDecf'] for entry in all_meta.values()]
all_pCO2 = [entry['Outcome']['pCO2'] for entry in all_meta.values()]
all_BE = [entry['Outcome']['BE'] for entry in all_meta.values()]
all_Apgar1 = [entry['Outcome']['Apgar1'] for entry in all_meta.values()]
all_Apgar5 = [entry['Outcome']['Apgar5'] for entry in all_meta.values()]
all_pH_lin = [10**(entry['Outcome']['pH']-7) for entry in all_meta.values()]
###Output
_____no_output_____
###Markdown
Show pH vs Apgar1
###Code
display_metric_vs_apgar(all_pH, all_Apgar1, limits=[6.5, 7.5],thresh=7,
title='pH vs Apgar1', xlabel='pH', ylabel='Apgar 1')
display_metric_vs_apgar(all_pH_lin, all_Apgar1, limits=[0,3], thresh=1,
title='linearized pH vs Apgar1', xlabel='linearized pH', ylabel='Apgar 1')
###Output
_____no_output_____
###Markdown
Show BDecf vs Apgar1
###Code
print(np.min(all_BDecf), np.max(all_BDecf))
display_metric_vs_apgar(all_BDecf, all_Apgar1, limits=[-5, 30], thresh=12,
title='BDecf vs Apgar1', xlabel='BDecf', ylabel='Apgar 1')
###Output
-3.4 26.11
###Markdown
Drilldown: BDecf vs Apgar1 for pH normal and abnormal pH
###Code
idx_subset = [i for i, x in enumerate(all_pH) if x <= 7]
display_metric_vs_apgar([all_BDecf[i] for i in idx_subset], [all_Apgar1[i] for i in idx_subset],
limits=[-5, 30], thresh=12,
title='pH vs Apgar1 when pH < 7', xlabel='BDecf', ylabel='Apgar 1')
idx_subset = [i for i, x in enumerate(all_pH) if x > 7]
display_metric_vs_apgar([all_BDecf[i] for i in idx_subset], [all_Apgar1[i] for i in idx_subset],
limits=[-5, 30], thresh=12,
title='pH vs Apgar1 when pH > 7', xlabel='BDecf', ylabel='Apgar 1')
###Output
_____no_output_____
###Markdown
Show BE vs Apgar1
###Code
print(np.min(all_BE), np.max(all_BE))
display_metric_vs_apgar(all_BE, all_Apgar1, limits=[-30, 0], thresh=-8,
title='BE vs Apgar1', xlabel='BE', ylabel='Apgar 1')
###Output
-26.8 -0.2
###Markdown
Drilldown: BE vs Apgar1 for pH normal and abnormal pH
###Code
idx_subset = [i for i, x in enumerate(all_pH) if x <= 7]
display_metric_vs_apgar([all_BE[i] for i in idx_subset], [all_Apgar1[i] for i in idx_subset],
limits=[-30, 0], thresh=-8,
title='BE vs Apgar1 when pH < 7', xlabel='BE', ylabel='Apgar 1')
idx_subset = [i for i, x in enumerate(all_pH) if x > 7]
display_metric_vs_apgar([all_BE[i] for i in idx_subset], [all_Apgar1[i] for i in idx_subset],
limits=[-30, 0], thresh=-8,
title='BE vs Apgar1 when pH > 7', xlabel='BE', ylabel='Apgar 1')
###Output
_____no_output_____
###Markdown
PCA
###Code
all_features = list(zip(all_Apgar1, all_BE, all_pH_lin, all_pH))
# Determine feature count by APGAR
counts = collections.defaultdict(int)
feature_by_apgar = collections.defaultdict(list)
for i, x in enumerate(all_Apgar1):
counts[x] += 1
feature_by_apgar[x].append(all_features[i])
for k in sorted(counts.keys()):
print(k, counts[k])
# equalize feature set without replace
max_entries = 14
train_features = []
train_labels = []
for k, v in feature_by_apgar.items():
if len(v) <= max_entries:
train_features += v
train_labels += [k] * len(v)
else:
train_labels += [k] * max_entries
for i in np.random.choice(np.arange(len(v)), size=max_entries, replace=False):
train_features.append(v[i])
train_features = np.vstack(train_features)
n_components = 1
pca = PCA(n_components=n_components)
pca.fit(train_features)
predict = pca.inverse_transform((pca.transform(train_features)))
display_predictions(train_labels, [x[0] for x in predict], title='predict vs Apgar1', xlabel='Apgar 1')
pca.components_
pca.explained_variance_ratio_
print(' all_Apgar1, all_BE, all_pH_lin all_pH')
print(pca.get_covariance())
###Output
all_Apgar1, all_BE, all_pH_lin all_pH
[[ 2.36527923 2.94616761 0.20461864 0.06199188]
[ 2.94616761 18.88614959 1.182813 0.35834858]
[ 0.20461864 1.182813 1.93776148 0.0248882 ]
[ 0.06199188 0.35834858 0.0248882 1.86315239]]
###Markdown
Try again ignoring pH (log version)
###Code
# strip pH from training set
train_features = [x[:-1] for x in train_features]
n_components = 1
pca = PCA(n_components=n_components)
pca.fit(train_features)
predict = pca.inverse_transform((pca.transform(train_features)))
display_predictions(train_labels, [x[0] for x in predict], title='predict vs Apgar1', xlabel='Apgar 1')
pca.components_
pca.explained_variance_ratio_
print(' all_Apgar1, all_BE, all_pH_lin')
print(pca.get_covariance())
###Output
all_Apgar1, all_BE, all_pH_lin
[[ 3.26447128 2.79037673 0.19376297]
[ 2.79037673 18.91680422 1.12040301]
[ 0.19376297 1.12040301 2.85970282]]
###Markdown
Modeling: Extimate Apgar1 using Outcome Metrics
###Code
all_features = list(zip(all_pH, all_BDecf, all_pCO2, all_BE, all_pH_lin))
###Output
_____no_output_____
###Markdown
SVM Regression 1Without Replacement (limit number of features by apgar)
###Code
# Determine feature count by APGAR
counts = collections.defaultdict(int)
feature_by_apgar1 = collections.defaultdict(list)
for i, x in enumerate(all_Apgar1):
counts[x] += 1
feature_by_apgar1[x].append(all_features[i])
for k in sorted(counts.keys()):
print(k, counts[k])
max_entries = 14
train_features = []
train_labels = []
for k, v in feature_by_apgar1.items():
print(k, len(v))
if len(v) <= max_entries:
train_features += v
train_labels += [k] * len(v)
else:
train_labels += [k] * max_entries
for i in np.random.choice(np.arange(len(v)), size=max_entries, replace=False):
train_features.append(v[i])
train_features = np.vstack(train_features)
train_features.shape, len(train_labels)
clf = svm.LinearSVR(max_iter=100000)
clf.fit(train_features, train_labels)
pred_train = clf.predict(train_features)
pred_test = clf.predict(all_features)
display_predictions(train_labels, pred_train, title='Training Apgar Predictions', xlabel='Apgar 1')
display_predictions(all_Apgar1, pred_test, title='All Apgar Predictions', xlabel='Apgar 1')
###Output
Error -- mean: 1.80 std: 1.23
###Markdown
SVM Regression 2With Replacement (feature set for sames features for each apgar score)
###Code
# Trtaining Set Using Replacement
max_entries = 50
train_features = []
train_labels = []
for k, v in feature_by_apgar1.items():
train_labels += [k] * max_entries
for i in np.random.choice(np.arange(len(v)), size=max_entries, replace=True):
train_features.append(v[i])
train_features = np.vstack(train_features)
train_features.shape, len(train_labels)
#clf = svm.LinearSVR(max_iter=100000)
clf = svm.SVR(max_iter=100000, gamma='scale')
clf.fit(train_features, train_labels)
pred_train = clf.predict(train_features)
pred_test = clf.predict(all_features)
display_predictions(all_Apgar1, pred_test, title='All Apgar Predictions', xlabel='Apgar 1')
###Output
Error -- mean: 1.92 std: 1.47
|
CNN scratch.ipynb | ###Markdown
Splitting Train, Validation, Test Data
###Code
train_dir = 'training_data'
val_dir = 'validation_data'
test_dir = 'test_data'
train_files = np.concatenate([cat_train, dog_train])
validate_files = np.concatenate([cat_val, dog_val])
test_files = np.concatenate([cat_test, dog_test])
os.mkdir(train_dir) if not os.path.isdir(train_dir) else None
os.mkdir(val_dir) if not os.path.isdir(val_dir) else None
os.mkdir(test_dir) if not os.path.isdir(test_dir) else None
for fn in train_files:
shutil.copy(fn, train_dir)
for fn in validate_files:
shutil.copy(fn, val_dir)
for fn in test_files:
shutil.copy(fn, test_dir)
#!rm -r test_data/ training_data/ validation_data/
from keras.preprocessing.image import ImageDataGenerator, load_img, img_to_array, array_to_img
IMG_DIM = (150,150)
train_files = glob.glob('training_data/*')
train_imgs = [];train_labels = []
for file in train_files:
try:
train_imgs.append( img_to_array(load_img( file,target_size=IMG_DIM )) )
train_labels.append(file.split('/')[1].split('_')[0])
except:
pass
train_imgs = np.array(train_imgs)
validation_files = glob.glob('validation_data/*')
validation_imgs = [];validation_labels = []
for file in validation_files:
try:
validation_imgs.append( img_to_array(load_img( file,target_size=IMG_DIM )) )
validation_labels.append(file.split('/')[1].split('_')[0])
except:
pass
train_imgs = np.array(train_imgs)
validation_imgs = np.array(validation_imgs)
print('Train dataset shape:', train_imgs.shape,
'\tValidation dataset shape:', validation_imgs.shape)
# encode text category labels
from sklearn.preprocessing import LabelEncoder
le = LabelEncoder()
le.fit(train_labels)
train_labels_enc = le.transform(train_labels)
validation_labels_enc = le.transform(validation_labels)
###Output
_____no_output_____
###Markdown
Image Augmentation
###Code
train_datagen = ImageDataGenerator(rescale=1./255,
zoom_range=0.3,
rotation_range=50,
width_shift_range=0.2,
height_shift_range=0.2,
shear_range=0.2,
horizontal_flip=True,
fill_mode='nearest')
val_datagen = ImageDataGenerator(rescale=1./255)
train_generator = train_datagen.flow(train_imgs, train_labels_enc, batch_size=30)
val_generator = val_datagen.flow(validation_imgs, validation_labels_enc, batch_size=20)
###Output
_____no_output_____
###Markdown
Keras Model
###Code
from keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout, Input
from keras.models import Model
from keras import optimizers
input_shape = (150, 150, 3)
input_l = Input((150,150,3))
l1_conv = Conv2D(16, kernel_size=(3, 3), activation='relu')(input_l)
l1_pool = MaxPooling2D(pool_size=(2, 2))(l1_conv)
l2_conv = Conv2D(64, kernel_size=(3, 3), activation='relu')(l1_pool)
l2_pool = MaxPooling2D(pool_size=(2, 2))(l2_conv)
l3_conv = Conv2D(128, kernel_size=(3, 3), activation='relu')(l2_pool)
l3_pool = MaxPooling2D(pool_size=(2, 2))(l3_conv)
l4 = Flatten()(l3_pool)
l4_dropout = Dropout(0.3)(l4)
l5 = Dense(512, activation='relu')(l4_dropout)
l5_dropout = Dropout(0.3)(l5)
output = Dense(1, activation='sigmoid')(l5_dropout)
model = Model(input_l, output)
model.compile(loss='binary_crossentropy', optimizer=optimizers.RMSprop(), metrics=['accuracy'])
model.summary()
history = model.fit_generator(train_generator, steps_per_epoch=100, epochs=100,
validation_data=val_generator, validation_steps=50,
verbose=2)
###Output
WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.
Instructions for updating:
Use tf.cast instead.
Epoch 1/100
- 17s - loss: 0.8622 - acc: 0.5310 - val_loss: 0.6788 - val_acc: 0.5240
Epoch 2/100
- 13s - loss: 0.6918 - acc: 0.5479 - val_loss: 0.6593 - val_acc: 0.5730
Epoch 3/100
- 13s - loss: 0.6772 - acc: 0.5933 - val_loss: 0.5895 - val_acc: 0.6900
Epoch 4/100
- 13s - loss: 0.6583 - acc: 0.6218 - val_loss: 0.6041 - val_acc: 0.6760
Epoch 5/100
- 13s - loss: 0.6430 - acc: 0.6293 - val_loss: 0.5745 - val_acc: 0.6990
Epoch 6/100
- 13s - loss: 0.6190 - acc: 0.6515 - val_loss: 0.5982 - val_acc: 0.6750
Epoch 7/100
- 14s - loss: 0.6172 - acc: 0.6640 - val_loss: 0.5325 - val_acc: 0.7120
Epoch 8/100
- 13s - loss: 0.6209 - acc: 0.6582 - val_loss: 0.5567 - val_acc: 0.7320
Epoch 9/100
- 13s - loss: 0.6211 - acc: 0.6560 - val_loss: 0.5010 - val_acc: 0.7410
Epoch 10/100
- 13s - loss: 0.5903 - acc: 0.6806 - val_loss: 0.5094 - val_acc: 0.7390
Epoch 11/100
- 13s - loss: 0.6082 - acc: 0.6700 - val_loss: 0.5429 - val_acc: 0.7120
Epoch 12/100
- 13s - loss: 0.5843 - acc: 0.6872 - val_loss: 0.5129 - val_acc: 0.7510
Epoch 13/100
- 14s - loss: 0.5839 - acc: 0.6810 - val_loss: 0.4997 - val_acc: 0.7640
Epoch 14/100
- 13s - loss: 0.5899 - acc: 0.6836 - val_loss: 0.5882 - val_acc: 0.7130
Epoch 15/100
- 13s - loss: 0.5805 - acc: 0.7000 - val_loss: 0.7053 - val_acc: 0.6550
Epoch 16/100
- 13s - loss: 0.5820 - acc: 0.6992 - val_loss: 0.4845 - val_acc: 0.7650
Epoch 17/100
- 13s - loss: 0.5752 - acc: 0.7017 - val_loss: 0.5363 - val_acc: 0.7270
Epoch 18/100
- 13s - loss: 0.5657 - acc: 0.7129 - val_loss: 0.4494 - val_acc: 0.7830
Epoch 19/100
- 14s - loss: 0.5787 - acc: 0.6917 - val_loss: 0.5283 - val_acc: 0.7290
Epoch 20/100
- 14s - loss: 0.5658 - acc: 0.7089 - val_loss: 0.4581 - val_acc: 0.7770
Epoch 21/100
- 14s - loss: 0.5553 - acc: 0.7147 - val_loss: 0.6111 - val_acc: 0.6850
Epoch 22/100
- 14s - loss: 0.5580 - acc: 0.7242 - val_loss: 0.5054 - val_acc: 0.7480
Epoch 23/100
- 13s - loss: 0.5588 - acc: 0.7100 - val_loss: 0.4720 - val_acc: 0.7730
Epoch 24/100
- 13s - loss: 0.5505 - acc: 0.7212 - val_loss: 0.5123 - val_acc: 0.7470
Epoch 25/100
- 15s - loss: 0.5712 - acc: 0.7073 - val_loss: 0.4787 - val_acc: 0.7610
Epoch 26/100
- 13s - loss: 0.5408 - acc: 0.7316 - val_loss: 0.4961 - val_acc: 0.7530
Epoch 27/100
- 13s - loss: 0.5300 - acc: 0.7393 - val_loss: 0.4686 - val_acc: 0.7830
Epoch 28/100
- 13s - loss: 0.5505 - acc: 0.7199 - val_loss: 0.5242 - val_acc: 0.7460
Epoch 29/100
- 13s - loss: 0.5271 - acc: 0.7457 - val_loss: 0.4545 - val_acc: 0.7900
Epoch 30/100
- 13s - loss: 0.5384 - acc: 0.7260 - val_loss: 0.4610 - val_acc: 0.7870
Epoch 31/100
- 14s - loss: 0.5223 - acc: 0.7367 - val_loss: 0.4424 - val_acc: 0.7960
Epoch 32/100
- 13s - loss: 0.5512 - acc: 0.7213 - val_loss: 0.4779 - val_acc: 0.7780
Epoch 33/100
- 13s - loss: 0.5463 - acc: 0.7433 - val_loss: 0.4244 - val_acc: 0.8020
Epoch 34/100
- 13s - loss: 0.5385 - acc: 0.7302 - val_loss: 0.5564 - val_acc: 0.7410
Epoch 35/100
- 13s - loss: 0.5427 - acc: 0.7340 - val_loss: 0.8722 - val_acc: 0.6250
Epoch 36/100
- 13s - loss: 0.5305 - acc: 0.7402 - val_loss: 0.4386 - val_acc: 0.8110
Epoch 37/100
- 14s - loss: 0.5184 - acc: 0.7457 - val_loss: 0.4450 - val_acc: 0.7910
Epoch 38/100
- 13s - loss: 0.5441 - acc: 0.7332 - val_loss: 0.4206 - val_acc: 0.8070
Epoch 39/100
- 13s - loss: 0.5285 - acc: 0.7407 - val_loss: 0.4497 - val_acc: 0.7910
Epoch 40/100
- 13s - loss: 0.5281 - acc: 0.7376 - val_loss: 0.4420 - val_acc: 0.8000
Epoch 41/100
- 13s - loss: 0.5231 - acc: 0.7410 - val_loss: 0.4106 - val_acc: 0.8160
Epoch 42/100
- 13s - loss: 0.5220 - acc: 0.7436 - val_loss: 0.5312 - val_acc: 0.7590
Epoch 43/100
- 15s - loss: 0.5149 - acc: 0.7517 - val_loss: 0.4484 - val_acc: 0.7830
Epoch 44/100
- 14s - loss: 0.5223 - acc: 0.7432 - val_loss: 0.4120 - val_acc: 0.8120
Epoch 45/100
- 14s - loss: 0.5195 - acc: 0.7400 - val_loss: 0.4429 - val_acc: 0.7960
Epoch 46/100
- 13s - loss: 0.5246 - acc: 0.7466 - val_loss: 0.4041 - val_acc: 0.8090
Epoch 47/100
- 13s - loss: 0.5075 - acc: 0.7553 - val_loss: 0.4265 - val_acc: 0.8250
Epoch 48/100
- 13s - loss: 0.5156 - acc: 0.7583 - val_loss: 0.4795 - val_acc: 0.7750
Epoch 49/100
- 14s - loss: 0.5120 - acc: 0.7517 - val_loss: 0.4283 - val_acc: 0.8020
Epoch 50/100
- 13s - loss: 0.5255 - acc: 0.7409 - val_loss: 0.4847 - val_acc: 0.7830
Epoch 51/100
- 13s - loss: 0.5193 - acc: 0.7490 - val_loss: 0.4578 - val_acc: 0.7940
Epoch 52/100
- 13s - loss: 0.5122 - acc: 0.7469 - val_loss: 0.4558 - val_acc: 0.8020
Epoch 53/100
- 13s - loss: 0.5049 - acc: 0.7603 - val_loss: 0.4554 - val_acc: 0.7800
Epoch 54/100
- 13s - loss: 0.5097 - acc: 0.7569 - val_loss: 0.4381 - val_acc: 0.7890
Epoch 55/100
- 15s - loss: 0.5139 - acc: 0.7637 - val_loss: 0.4017 - val_acc: 0.8140
Epoch 56/100
- 13s - loss: 0.5099 - acc: 0.7560 - val_loss: 0.4347 - val_acc: 0.7940
Epoch 57/100
- 13s - loss: 0.4980 - acc: 0.7530 - val_loss: 0.4394 - val_acc: 0.7860
Epoch 58/100
- 13s - loss: 0.5130 - acc: 0.7589 - val_loss: 0.4424 - val_acc: 0.7960
Epoch 59/100
- 13s - loss: 0.4981 - acc: 0.7733 - val_loss: 0.4552 - val_acc: 0.7710
Epoch 60/100
- 13s - loss: 0.4991 - acc: 0.7599 - val_loss: 0.4227 - val_acc: 0.8080
Epoch 61/100
- 15s - loss: 0.4875 - acc: 0.7707 - val_loss: 0.4253 - val_acc: 0.8080
Epoch 62/100
- 13s - loss: 0.4981 - acc: 0.7632 - val_loss: 0.5508 - val_acc: 0.7510
Epoch 63/100
- 13s - loss: 0.5062 - acc: 0.7603 - val_loss: 0.4149 - val_acc: 0.8180
Epoch 64/100
- 13s - loss: 0.5008 - acc: 0.7655 - val_loss: 0.3925 - val_acc: 0.8360
Epoch 65/100
- 14s - loss: 0.4924 - acc: 0.7760 - val_loss: 0.4087 - val_acc: 0.8190
Epoch 66/100
- 13s - loss: 0.4925 - acc: 0.7642 - val_loss: 0.4290 - val_acc: 0.8010
Epoch 67/100
- 15s - loss: 0.4722 - acc: 0.7770 - val_loss: 0.3828 - val_acc: 0.8220
Epoch 68/100
- 14s - loss: 0.5055 - acc: 0.7606 - val_loss: 0.4122 - val_acc: 0.8120
Epoch 69/100
- 13s - loss: 0.4900 - acc: 0.7737 - val_loss: 0.4063 - val_acc: 0.8290
Epoch 70/100
- 13s - loss: 0.4993 - acc: 0.7636 - val_loss: 0.4151 - val_acc: 0.8010
Epoch 71/100
- 13s - loss: 0.5020 - acc: 0.7690 - val_loss: 0.4158 - val_acc: 0.7990
Epoch 72/100
- 13s - loss: 0.4955 - acc: 0.7653 - val_loss: 0.4049 - val_acc: 0.8310
Epoch 73/100
- 14s - loss: 0.4823 - acc: 0.7830 - val_loss: 0.4336 - val_acc: 0.8050
Epoch 74/100
- 13s - loss: 0.4804 - acc: 0.7819 - val_loss: 0.3934 - val_acc: 0.8180
Epoch 75/100
- 13s - loss: 0.5065 - acc: 0.7643 - val_loss: 0.4974 - val_acc: 0.7810
Epoch 76/100
- 13s - loss: 0.4888 - acc: 0.7779 - val_loss: 0.4695 - val_acc: 0.7960
Epoch 77/100
- 13s - loss: 0.4895 - acc: 0.7730 - val_loss: 0.4161 - val_acc: 0.8160
Epoch 78/100
- 13s - loss: 0.4820 - acc: 0.7790 - val_loss: 0.3894 - val_acc: 0.8220
Epoch 79/100
- 14s - loss: 0.4907 - acc: 0.7703 - val_loss: 0.3769 - val_acc: 0.8350
Epoch 80/100
- 13s - loss: 0.4794 - acc: 0.7779 - val_loss: 0.4107 - val_acc: 0.8180
Epoch 81/100
- 13s - loss: 0.4800 - acc: 0.7823 - val_loss: 0.4767 - val_acc: 0.7850
Epoch 82/100
- 13s - loss: 0.4937 - acc: 0.7699 - val_loss: 0.4065 - val_acc: 0.8190
Epoch 83/100
- 13s - loss: 0.4650 - acc: 0.7860 - val_loss: 0.3732 - val_acc: 0.8380
Epoch 84/100
- 13s - loss: 0.4981 - acc: 0.7692 - val_loss: 0.4256 - val_acc: 0.8300
Epoch 85/100
- 14s - loss: 0.5049 - acc: 0.7670 - val_loss: 0.3864 - val_acc: 0.8290
Epoch 86/100
- 13s - loss: 0.4730 - acc: 0.7830 - val_loss: 0.4212 - val_acc: 0.8200
Epoch 87/100
- 13s - loss: 0.4796 - acc: 0.7730 - val_loss: 0.4310 - val_acc: 0.8160
Epoch 88/100
- 14s - loss: 0.4944 - acc: 0.7702 - val_loss: 0.3845 - val_acc: 0.8290
Epoch 89/100
- 13s - loss: 0.4922 - acc: 0.7790 - val_loss: 0.3772 - val_acc: 0.8360
Epoch 90/100
- 13s - loss: 0.4921 - acc: 0.7816 - val_loss: 0.3902 - val_acc: 0.8320
Epoch 91/100
- 15s - loss: 0.4711 - acc: 0.7860 - val_loss: 0.4081 - val_acc: 0.8130
Epoch 92/100
- 13s - loss: 0.5064 - acc: 0.7709 - val_loss: 0.3651 - val_acc: 0.8450
Epoch 93/100
- 13s - loss: 0.4929 - acc: 0.7687 - val_loss: 0.3481 - val_acc: 0.8420
Epoch 94/100
- 13s - loss: 0.4720 - acc: 0.7796 - val_loss: 0.4129 - val_acc: 0.8060
Epoch 95/100
- 13s - loss: 0.4835 - acc: 0.7820 - val_loss: 0.5525 - val_acc: 0.7710
Epoch 96/100
- 13s - loss: 0.4742 - acc: 0.7709 - val_loss: 0.4102 - val_acc: 0.8290
Epoch 97/100
- 15s - loss: 0.4675 - acc: 0.7807 - val_loss: 0.5125 - val_acc: 0.7810
Epoch 98/100
- 13s - loss: 0.4692 - acc: 0.7786 - val_loss: 0.3939 - val_acc: 0.8340
Epoch 99/100
- 13s - loss: 0.4961 - acc: 0.7790 - val_loss: 0.4290 - val_acc: 0.8240
Epoch 100/100
- 13s - loss: 0.4662 - acc: 0.7876 - val_loss: 0.4970 - val_acc: 0.7820
###Markdown
Model Performance
###Code
%matplotlib inline
import matplotlib.pyplot as plt
f, (ax1, ax2) = plt.subplots(1, 2, figsize=(12, 4))
t = f.suptitle('Basic CNN Performance', fontsize=12)
f.subplots_adjust(top=0.85, wspace=0.3)
epoch_list = list(range(1,101))
ax1.plot(epoch_list, history.history['acc'], label='Train Accuracy')
ax1.plot(epoch_list, history.history['val_acc'], label='Validation Accuracy')
ax1.set_xticks(np.arange(0, 101, 5))
ax1.set_ylabel('Accuracy Value')
ax1.set_xlabel('Epoch')
ax1.set_title('Accuracy')
l1 = ax1.legend(loc="best")
ax2.plot(epoch_list, history.history['loss'], label='Train Loss')
ax2.plot(epoch_list, history.history['val_loss'], label='Validation Loss')
ax2.set_xticks(np.arange(0, 101, 5))
ax2.set_ylabel('Loss Value')
ax2.set_xlabel('Epoch')
ax2.set_title('Loss')
l2 = ax2.legend(loc="best")
if not os.path.exists('saved_models'): os.mkdir('saved_models')
model.save('models/cnn scratch.h5')
###Output
_____no_output_____ |
vae_coordconv.ipynb | ###Markdown
Variational Autoencoder CoordConv filters
###Code
from keras.utils.vis_utils import plot_model
from keras_visualizer import visualizer
from keras.callbacks import EarlyStopping, TensorBoard
from keras.preprocessing.image import ImageDataGenerator
import numpy as np
import matplotlib.pyplot as plt
from autoencoders.VAE import VAECoordConv as VAE
DATASET_SIZE = 25084
INPUT_SHAPE = (40, 40, 1)
import tensorflow as tf
devices = tf.config.list_physical_devices('GPU')
if len(devices) < 1:
raise Exception("Cannot initialize GPU")
print("GPU configured correctly")
datagen = ImageDataGenerator(
rescale=1./255,
validation_split=0.3,
horizontal_flip=True,
vertical_flip=True
)
# Allow horizontal flip and vertical flip as a mirror image in both axes of a game is a valid game state
train_datagen = datagen.flow_from_directory(
'images_trans/',
target_size=(INPUT_SHAPE[0], INPUT_SHAPE[1]),
color_mode='grayscale',
class_mode='input',
shuffle=True,
subset='training'
)
val_datagen = datagen.flow_from_directory(
'images_trans/',
target_size=(INPUT_SHAPE[0], INPUT_SHAPE[1]),
color_mode='grayscale',
class_mode='input',
shuffle=True,
subset='validation'
)
vae = VAE(
layers=5,
latent_size=16,
kernel_size=3,
input_shape=INPUT_SHAPE,
filters=16,
name="VAE"
)
vae.summary()
plot_model(vae.decoder, to_file='coord_dec.png', show_shapes=True)
callbacks = [
EarlyStopping(monitor='val_loss', patience=40),
TensorBoard(
log_dir='./logs',
histogram_freq=1
)
]
history = vae.train(
train_datagen,
val_datagen,
epochs=75,
callbacks=callbacks
)
val_rec_loss = history.history['val_reconstruction_loss']
rec_loss = history.history['reconstruction_loss']
epochs = range(1, len(val_rec_loss) + 1)
plt.figure(figsize=(8,6), dpi=80)
plt.plot(epochs, val_rec_loss, 'r', label='Val Reconstruction loss')
plt.plot(epochs, rec_loss, 'b', label='Reconstruction loss')
plt.title('CoordConv training')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
from keras.preprocessing.image import load_img
from keras.preprocessing.image import img_to_array
import random
n = 10
images = np.empty((n, *INPUT_SHAPE))
for i in range(n):
rand_img = random.randint(0, DATASET_SIZE)
img = img_to_array(load_img(f"images_trans/pong_trans_{rand_img}.png", color_mode='grayscale'))
images[i] = img
decoded_imgs = vae.predict(images)
latent = vae.encoder.predict(images)
#print(latent[:][2])
decoded_imgs = vae.decoder.predict(latent[:][2])
plt.figure(figsize=(20, 4))
for i in range(1, n+1):
# Display original
ax = plt.subplot(2, n, i)
plt.imshow(images[i-1].reshape(INPUT_SHAPE[0], INPUT_SHAPE[1]))
plt.gray()
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
# Display reconstruction
ax = plt.subplot(2, n, i + n)
plt.imshow(decoded_imgs[i-1].reshape(INPUT_SHAPE[0], INPUT_SHAPE[1]))
plt.gray()
ax.get_xaxis().set_visible(False)
ax.get_yaxis().set_visible(False)
plt.show()
import random
def plot_latent_space(vae, n=15, figsize=25):
# display a n*n 2D manifold of digits
x_input_size = INPUT_SHAPE[0]
y_input_size = INPUT_SHAPE[1]
scale = 100.0
figure = np.zeros((x_input_size * n + n - 1, y_input_size * n + n - 1))
# linearly spaced coordinates corresponding to the 2D plot
# of digit classes in the latent space
grid_x = np.linspace(-scale, scale, n)
grid_y = np.linspace(-scale, scale, n)[::-1]
for i, yi in enumerate(grid_y):
for j, xi in enumerate(grid_x):
rest = rest = [i * 10 for i in range(-7, 7)]
# missing = -2.9679005e+00, -1.7151514e+01
z_sample = np.array([[xi, yi] + rest])
#z_sample = np.array([[random.randint(-scale, scale) for i in range(16)]])
x_decoded = vae.decoder.predict(z_sample)
digit = x_decoded[0].reshape(x_input_size, y_input_size)
figure[
i * x_input_size + i: (i + 1) * x_input_size + i,
j * y_input_size + j: (j + 1) * y_input_size + j,
] = digit
if i != len(grid_x) - 1:
figure[
(i + 1) * x_input_size + i,
j * y_input_size + j: (j + 1) * y_input_size + j,
] = np.array([1] * y_input_size)
if j != len(grid_y) - 1:
figure[
i * x_input_size + i: (i + 1) * x_input_size + i,
(j + 1) * y_input_size + j,
] = np.array([1] * x_input_size)
plt.figure(figsize=(figsize, figsize))
plt.imshow(figure, cmap="Greys_r")
plt.show()
plot_latent_space(vae)
###Output
_____no_output_____ |
Issues/algorithms/Bit Manipulation.ipynb | ###Markdown
5.1 Insertion:You are given two 32-bit numbers, N and M, and two bit positions, i and j. Write a methodto insert M into N such that M starts at bit j and ends at bit i. You can assume that the bits j throughi have enough space to fit all of M. That is, if M = 10011, you can assume that there are at least 5bits between j and i. You would not, for example, have j = 3 and i = 2, because M could not fullyfit between bit 3 and bit 2.
###Code
bit = 0xFFFF
M = 0x5
N = 0xF00F
def insertion(M, N, i, j):
ALL = 0xFFFFFFFF
left_mask = ALL << j + 1
print(bin(ALL))
print("left_mask:" + bin(left_mask))
right_mask = ALL >> (32 - i)
mask = left_mask|right_mask
print(bin(mask))
M_ = (M << i)
ret = (N&mask) | M_
print(bin(M_))
print(bin(ret))
return ret
print(bin(insertion(M, N, 3, 5)))
###Output
0b11111111111111111111111111111111
left_mask:0b11111111111111111111111111111111000000
0b11111111111111111111111111111111000111
0b101000
0b1111000000101111
0b1111000000101111
###Markdown
5.2 Binary to String: Given a real number between 0 and 1 (e.g., 0.72) that is passed in as a double, print the binary representation. If the number cannot be represented accurately in binary with at most 32 characters, print"ERROR:'
###Code
def bin_to_str(number):
strs = []
remain = number
for i in range(1, 31):
if remain == 0:
break
elif remain - 0.5**i >= 0:
remain = remain - 0.5**i
strs.append(str(1))
else:
strs.append(str(0))
print(remain, 0.5**i)
if remain > 0:
return "ERROR"
else:
return "0." + "".join(strs)
print(bin_to_str(0.875))
###Output
0.375 0.5
0.125 0.25
0.0 0.125
0.111
|
_notebooks/2022-03-15-Assignment07.ipynb | ###Markdown
"Stats & Modeling with Romeo and Juliet"> "This is assignment 7 from DH 140 with Professor Benjamin Winjum at UCLA for Winter Quarter 2022. It employs methods to run analysis on the play and output some exploratory analysis. The second part of the assignment look at a test dataset to introduce how to employ a Linear Regression model using the dataframes provided and ways to select a good feature for model training."- toc:true- branch: master- badges: true- comments: true- author: Anh Mac- categories: [fastpages, jupyter, Shakespeare, machinelearning] Shakespeare play from Week 6- Tokenize the words, remove stopwords, stem or lemmatize the words, and calculate the word frequencies- For the word frequencies, calculate the mean, median, mode, and trimmed mean.- For the trimmed mean, you can choose what to trim, but comment on the number used for trimming.- Plot a histogram of the word frequency data and comment on the relative locations of the mean, median, mode, and trimmed mean- Calculate the standard deviation and the interquartile range (the difference of the 75% and 25% quantile)- Comment as well on how they compare to each other and to the histogram plot.
###Code
#hide
import nltk
from nltk.tokenize import word_tokenize, sent_tokenize
nltk.download('punkt')
# Read the file that contains text of Romeo and Julie
f = open("romeo-and-juliet.txt", "r")
text = f.read()
###Output
_____no_output_____
###Markdown
Tokenize words
###Code
# Tokenize the words
sent = sent_tokenize(text)
print(word_tokenize(sent[1]))
words = []
for s in sent:
for w in word_tokenize(s):
words.append(w)
###Output
_____no_output_____
###Markdown
Remove stopwords
###Code
# remove stopwords
from nltk.corpus import stopwords
from string import punctuation
nltk.download('stopwords')
myStopWords = list(punctuation) + stopwords.words('english') + ['’']
wordsNoStop = [w for w in words if w not in myStopWords]
###Output
_____no_output_____
###Markdown
Stem the words
###Code
#hide
from nltk.stem.lancaster import LancasterStemmer
wordLancasterStems = [LancasterStemmer().stem(w) for w in wordsNoStop]
###Output
_____no_output_____
###Markdown
Calculate frequencies
###Code
#hide
import collections
wordCount = collections.Counter(wordLancasterStems)
frequencies = wordCount.most_common()
frequencies[0:5]
###Output
_____no_output_____
###Markdown
Calculate mean/median/mode/trimmed mean
###Code
import pandas as pd
df = pd.DataFrame(frequencies, columns=["text", "frequency"])
#collapse-hide
print("MEAN: ", df["frequency"].mean())
print("MEDIAN:", df["frequency"].median())
print("MODE: ", df["frequency"].mode())
df2 = df.sort_values(by="frequency",ignore_index=True).copy()
df2
###Output
_____no_output_____
###Markdown
I chose to trim 20% of the data since the outliers look like it's only the higher numbers like 314 and 647, and there also seems to be a lot of 1's, so remove some would not skewed our data too much. I mostly want to remove the higher outliers.
###Code
print("TRIMMED MEAN:", df2.loc[int(0.2*3053):int(0.8*3053),'frequency'].mean())
###Output
TRIMMED MEAN: 2.0845608292416804
###Markdown
Histogram
###Code
df["frequency"].plot.hist(title='Words frequencies in Romeo & Juliet')
###Output
_____no_output_____
###Markdown
**Comment on histogram:**- Most of the values are 1, so the Mode being 1 makes sense.- Median of 2 also makes sense as most of the data seems to be between 0-50, and as we can see from a preview of the dataframe that most of these numbers would be between 1-2, and the middle number should be around there as well.- The original mean of 6.188932547478716 seems a bit high when looking at the distribution, so our trimmed mean of 2.0845608292416804 seems to make more sense with the outliers above 300 removed. Standard Deviation & Interquartile range
###Code
print("STD: ", df["frequency"].std())
df["frequency"].quantile(0.75) - df["frequency"].quantile(0.25)
###Output
_____no_output_____
###Markdown
**Comment on std and interquartile range:**- The STD of 19.957504689286747 makes sense considering our small Mode, Median, and larger outliers to the right. The standard deviation calculates the distance between every data point so the large outliers skewed this number to be larger.- The interquartile range of 3.0 also makes sense as our most of our data has smaller frequency, and the plot is heavily skewed left, so the interquartile range signifies that most of our data are close together in distance. The interquartile range calculate the distance between the 50% middle data points, so in terms of the distribution of our histogram, this value makes sense. Foray into machine learning- Import scikit-learn's example diabetes dataset as a Panda's dataframe with the following code:from sklearn import datasetsdf = datasets.load_diabetes(as_frame=True) features_df = df.datatarget_df = df.target- Use the following code to view a description of the dataset:print(df.DESCR)- Do some exploratory data analysis of the features, including getting summary statistical information- Find the column in features_df that has the highest correlation coefficient with the target values in target_df- Make a scatter plot of the target values vs this feature column's values and comment on how the plotted points match up with the correlation coefficient- Using this feature and target, perform linear regression with sklearn's LinearRegression- Print the coefficients of the model- Plot the linear fit on top of the scatter plot- Calculate (or output) the mean squared error and R-squared values for your fit- Try doing linear regression with another variable and check how the new fit's mean squared error and R-squared values change.
###Code
from sklearn import datasets
df = datasets.load_diabetes(as_frame=True)
features_df = df.data
target_df = df.target
print(df.DESCR)
###Output
.. _diabetes_dataset:
Diabetes dataset
----------------
Ten baseline variables, age, sex, body mass index, average blood
pressure, and six blood serum measurements were obtained for each of n =
442 diabetes patients, as well as the response of interest, a
quantitative measure of disease progression one year after baseline.
**Data Set Characteristics:**
:Number of Instances: 442
:Number of Attributes: First 10 columns are numeric predictive values
:Target: Column 11 is a quantitative measure of disease progression one year after baseline
:Attribute Information:
- age age in years
- sex
- bmi body mass index
- bp average blood pressure
- s1 tc, total serum cholesterol
- s2 ldl, low-density lipoproteins
- s3 hdl, high-density lipoproteins
- s4 tch, total cholesterol / HDL
- s5 ltg, possibly log of serum triglycerides level
- s6 glu, blood sugar level
Note: Each of these 10 feature variables have been mean centered and scaled by the standard deviation times `n_samples` (i.e. the sum of squares of each column totals 1).
Source URL:
https://www4.stat.ncsu.edu/~boos/var.select/diabetes.html
For more information see:
Bradley Efron, Trevor Hastie, Iain Johnstone and Robert Tibshirani (2004) "Least Angle Regression," Annals of Statistics (with discussion), 407-499.
(https://web.stanford.edu/~hastie/Papers/LARS/LeastAngle_2002.pdf)
###Markdown
Exploratory Data Analysis
###Code
features_df.describe()
features_df[features_df.columns[:]].hist(figsize=(20, 20))
###Output
_____no_output_____
###Markdown
Find column with highest correlation
###Code
col=list(features_df.columns)
df2=features_df.copy()
df2["target"] = list(target_df.values)
for c in col:
print(df2[[c,'target']].corr())
###Output
age target
age 1.000000 0.187889
target 0.187889 1.000000
sex target
sex 1.000000 0.043062
target 0.043062 1.000000
bmi target
bmi 1.00000 0.58645
target 0.58645 1.00000
bp target
bp 1.000000 0.441484
target 0.441484 1.000000
s1 target
s1 1.000000 0.212022
target 0.212022 1.000000
s2 target
s2 1.000000 0.174054
target 0.174054 1.000000
s3 target
s3 1.000000 -0.394789
target -0.394789 1.000000
s4 target
s4 1.000000 0.430453
target 0.430453 1.000000
s5 target
s5 1.000000 0.565883
target 0.565883 1.000000
s6 target
s6 1.000000 0.382483
target 0.382483 1.000000
###Markdown
Column with highest correlation to **target_df** is **bmi**. Scatter plot
###Code
df2.plot.scatter(y='target',x='bmi')
###Output
_____no_output_____
###Markdown
**Comment on scatter plot:**- The plotted points does show somewhat of a linear trend upward that matches up with the correlation coefficient.- We can vaguely observe a diagonal line through the points Linear Regression
###Code
import numpy as np
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error, r2_score
X = np.array(df2['bmi'])
y = np.array(df2['target'])
reg = LinearRegression().fit(X.reshape(-1, 1), y)
###Output
_____no_output_____
###Markdown
Coefficient
###Code
coef = reg.coef_
coef
###Output
_____no_output_____
###Markdown
Linear fit
###Code
ytrain = reg.intercept_ + reg.coef_ * X
plt.plot(X,y,'ro',X,ytrain,'b-');
###Output
_____no_output_____
###Markdown
Calculate Mean-Squared-Error and R^2
###Code
mean_squared_error(y, ytrain)
r2_score(y, ytrain)
###Output
_____no_output_____
###Markdown
Linear Regression with s5
###Code
df2.plot.scatter(y='target',x='s5')
X = np.array(df2['s5'])
y = np.array(df2['target'])
reg = LinearRegression().fit(X.reshape(-1, 1), y)
ytrain = reg.intercept_ + reg.coef_ * X
plt.plot(X,y,'ro',X,ytrain,'b-');
mean_squared_error(y, ytrain)
r2_score(y, ytrain)
###Output
_____no_output_____ |
example/03_peak-search/peak-search-python.ipynb | ###Markdown
How to find peaks of impedance curve It is quite easy using scipy.signal .
###Code
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
# module to do peak search
from scipy.signal import argrelextrema
# read data
dd = pd.read_csv('side-hole.imp',header=0)
###Output
_____no_output_____
###Markdown
This impedance curve is a weird to do peak search due to side-hole.
###Code
p1 = dd.plot(x = 'freq', y = 'imp.mag')
###Output
_____no_output_____
###Markdown
Pick up ndarray from dataframe to use argrelextrema.
###Code
y = dd.iloc[:,3].values
type(y)
###Output
_____no_output_____
###Markdown
Find peaks location. Easy.
###Code
idx = argrelextrema(y, np.greater);
idx
y[idx]
x = dd.iloc[:,0].values
x[idx]
###Output
_____no_output_____
###Markdown
To super impose peaks, using normal matplot function is better.
###Code
plt.scatter(x[idx],y[idx],color='r')
plt.plot(x,y)
###Output
_____no_output_____
###Markdown
In case of searching local minima, use np.less.
###Code
i2 = argrelextrema(y,np.less)
x2 = x[i2]
y2 = y[i2]
plt.plot(x,y)
plt.scatter(x2,y2,s = 50,color = 'green', marker = '+')
###Output
_____no_output_____ |
experiments/baseline_ptn_32bit/oracle.run1/trials/0/trial.ipynb | ###Markdown
PTN TemplateThis notebook serves as a template for single dataset PTN experiments It can be run on its own by setting STANDALONE to True (do a find for "STANDALONE" to see where) But it is intended to be executed as part of a *papermill.py script. See any of the experimentes with a papermill script to get started with that workflow.
###Code
%load_ext autoreload
%autoreload 2
%matplotlib inline
import os, json, sys, time, random
import numpy as np
import torch
from torch.optim import Adam
from easydict import EasyDict
import matplotlib.pyplot as plt
from steves_models.steves_ptn import Steves_Prototypical_Network
from steves_utils.lazy_iterable_wrapper import Lazy_Iterable_Wrapper
from steves_utils.iterable_aggregator import Iterable_Aggregator
from steves_utils.ptn_train_eval_test_jig import PTN_Train_Eval_Test_Jig
from steves_utils.torch_sequential_builder import build_sequential
from steves_utils.torch_utils import get_dataset_metrics, ptn_confusion_by_domain_over_dataloader
from steves_utils.utils_v2 import (per_domain_accuracy_from_confusion, get_datasets_base_path)
from steves_utils.PTN.utils import independent_accuracy_assesment
from steves_utils.stratified_dataset.episodic_accessor import Episodic_Accessor_Factory
from steves_utils.ptn_do_report import (
get_loss_curve,
get_results_table,
get_parameters_table,
get_domain_accuracies,
)
from steves_utils.transforms import get_chained_transform
###Output
_____no_output_____
###Markdown
Required ParametersThese are allowed parameters, not defaultsEach of these values need to be present in the injected parameters (the notebook will raise an exception if they are not present)Papermill uses the cell tag "parameters" to inject the real parameters below this cell.Enable tags to see what I mean
###Code
required_parameters = {
"experiment_name",
"lr",
"device",
"seed",
"dataset_seed",
"labels_source",
"labels_target",
"domains_source",
"domains_target",
"num_examples_per_domain_per_label_source",
"num_examples_per_domain_per_label_target",
"n_shot",
"n_way",
"n_query",
"train_k_factor",
"val_k_factor",
"test_k_factor",
"n_epoch",
"patience",
"criteria_for_best",
"x_transforms_source",
"x_transforms_target",
"episode_transforms_source",
"episode_transforms_target",
"pickle_name",
"x_net",
"NUM_LOGS_PER_EPOCH",
"BEST_MODEL_PATH",
"torch_default_dtype"
}
standalone_parameters = {}
standalone_parameters["experiment_name"] = "STANDALONE PTN"
standalone_parameters["lr"] = 0.0001
standalone_parameters["device"] = "cuda"
standalone_parameters["seed"] = 1337
standalone_parameters["dataset_seed"] = 1337
standalone_parameters["num_examples_per_domain_per_label_source"]=100
standalone_parameters["num_examples_per_domain_per_label_target"]=100
standalone_parameters["n_shot"] = 3
standalone_parameters["n_query"] = 2
standalone_parameters["train_k_factor"] = 1
standalone_parameters["val_k_factor"] = 2
standalone_parameters["test_k_factor"] = 2
standalone_parameters["n_epoch"] = 100
standalone_parameters["patience"] = 10
standalone_parameters["criteria_for_best"] = "target_accuracy"
standalone_parameters["x_transforms_source"] = ["unit_power"]
standalone_parameters["x_transforms_target"] = ["unit_power"]
standalone_parameters["episode_transforms_source"] = []
standalone_parameters["episode_transforms_target"] = []
standalone_parameters["torch_default_dtype"] = "torch.float32"
standalone_parameters["x_net"] = [
{"class": "nnReshape", "kargs": {"shape":[-1, 1, 2, 256]}},
{"class": "Conv2d", "kargs": { "in_channels":1, "out_channels":256, "kernel_size":(1,7), "bias":False, "padding":(0,3), },},
{"class": "ReLU", "kargs": {"inplace": True}},
{"class": "BatchNorm2d", "kargs": {"num_features":256}},
{"class": "Conv2d", "kargs": { "in_channels":256, "out_channels":80, "kernel_size":(2,7), "bias":True, "padding":(0,3), },},
{"class": "ReLU", "kargs": {"inplace": True}},
{"class": "BatchNorm2d", "kargs": {"num_features":80}},
{"class": "Flatten", "kargs": {}},
{"class": "Linear", "kargs": {"in_features": 80*256, "out_features": 256}}, # 80 units per IQ pair
{"class": "ReLU", "kargs": {"inplace": True}},
{"class": "BatchNorm1d", "kargs": {"num_features":256}},
{"class": "Linear", "kargs": {"in_features": 256, "out_features": 256}},
]
# Parameters relevant to results
# These parameters will basically never need to change
standalone_parameters["NUM_LOGS_PER_EPOCH"] = 10
standalone_parameters["BEST_MODEL_PATH"] = "./best_model.pth"
# uncomment for CORES dataset
from steves_utils.CORES.utils import (
ALL_NODES,
ALL_NODES_MINIMUM_1000_EXAMPLES,
ALL_DAYS
)
standalone_parameters["labels_source"] = ALL_NODES
standalone_parameters["labels_target"] = ALL_NODES
standalone_parameters["domains_source"] = [1]
standalone_parameters["domains_target"] = [2,3,4,5]
standalone_parameters["pickle_name"] = "cores.stratified_ds.2022A.pkl"
# Uncomment these for ORACLE dataset
# from steves_utils.ORACLE.utils_v2 import (
# ALL_DISTANCES_FEET,
# ALL_RUNS,
# ALL_SERIAL_NUMBERS,
# )
# standalone_parameters["labels_source"] = ALL_SERIAL_NUMBERS
# standalone_parameters["labels_target"] = ALL_SERIAL_NUMBERS
# standalone_parameters["domains_source"] = [8,20, 38,50]
# standalone_parameters["domains_target"] = [14, 26, 32, 44, 56]
# standalone_parameters["pickle_name"] = "oracle.frame_indexed.stratified_ds.2022A.pkl"
# standalone_parameters["num_examples_per_domain_per_label_source"]=1000
# standalone_parameters["num_examples_per_domain_per_label_target"]=1000
# Uncomment these for Metahan dataset
# standalone_parameters["labels_source"] = list(range(19))
# standalone_parameters["labels_target"] = list(range(19))
# standalone_parameters["domains_source"] = [0]
# standalone_parameters["domains_target"] = [1]
# standalone_parameters["pickle_name"] = "metehan.stratified_ds.2022A.pkl"
# standalone_parameters["n_way"] = len(standalone_parameters["labels_source"])
# standalone_parameters["num_examples_per_domain_per_label_source"]=200
# standalone_parameters["num_examples_per_domain_per_label_target"]=100
standalone_parameters["n_way"] = len(standalone_parameters["labels_source"])
# Parameters
parameters = {
"experiment_name": "baseline_ptn_32bit_oracle.run1",
"lr": 0.001,
"device": "cuda",
"seed": 1337,
"dataset_seed": 1337,
"labels_source": [
"3123D52",
"3123D65",
"3123D79",
"3123D80",
"3123D54",
"3123D70",
"3123D7B",
"3123D89",
"3123D58",
"3123D76",
"3123D7D",
"3123EFE",
"3123D64",
"3123D78",
"3123D7E",
"3124E4A",
],
"labels_target": [
"3123D52",
"3123D65",
"3123D79",
"3123D80",
"3123D54",
"3123D70",
"3123D7B",
"3123D89",
"3123D58",
"3123D76",
"3123D7D",
"3123EFE",
"3123D64",
"3123D78",
"3123D7E",
"3124E4A",
],
"x_transforms_source": [],
"x_transforms_target": [],
"episode_transforms_source": [],
"episode_transforms_target": [],
"num_examples_per_domain_per_label_source": 1000,
"num_examples_per_domain_per_label_target": 1000,
"n_shot": 3,
"n_way": 16,
"n_query": 2,
"train_k_factor": 1,
"val_k_factor": 2,
"test_k_factor": 2,
"torch_default_dtype": "torch.float32",
"n_epoch": 50,
"patience": 3,
"criteria_for_best": "target_loss",
"x_net": [
{"class": "nnReshape", "kargs": {"shape": [-1, 1, 2, 256]}},
{
"class": "Conv2d",
"kargs": {
"in_channels": 1,
"out_channels": 256,
"kernel_size": [1, 7],
"bias": False,
"padding": [0, 3],
},
},
{"class": "ReLU", "kargs": {"inplace": True}},
{"class": "BatchNorm2d", "kargs": {"num_features": 256}},
{
"class": "Conv2d",
"kargs": {
"in_channels": 256,
"out_channels": 80,
"kernel_size": [2, 7],
"bias": True,
"padding": [0, 3],
},
},
{"class": "ReLU", "kargs": {"inplace": True}},
{"class": "BatchNorm2d", "kargs": {"num_features": 80}},
{"class": "Flatten", "kargs": {}},
{"class": "Linear", "kargs": {"in_features": 20480, "out_features": 256}},
{"class": "ReLU", "kargs": {"inplace": True}},
{"class": "BatchNorm1d", "kargs": {"num_features": 256}},
{"class": "Linear", "kargs": {"in_features": 256, "out_features": 256}},
],
"NUM_LOGS_PER_EPOCH": 10,
"BEST_MODEL_PATH": "./best_model.pth",
"pickle_name": "oracle.Run1_10kExamples_stratified_ds.2022A.pkl",
"domains_source": [8, 32, 50],
"domains_target": [14, 20, 26, 38, 44],
}
# Set this to True if you want to run this template directly
STANDALONE = False
if STANDALONE:
print("parameters not injected, running with standalone_parameters")
parameters = standalone_parameters
if not 'parameters' in locals() and not 'parameters' in globals():
raise Exception("Parameter injection failed")
#Use an easy dict for all the parameters
p = EasyDict(parameters)
supplied_keys = set(p.keys())
if supplied_keys != required_parameters:
print("Parameters are incorrect")
if len(supplied_keys - required_parameters)>0: print("Shouldn't have:", str(supplied_keys - required_parameters))
if len(required_parameters - supplied_keys)>0: print("Need to have:", str(required_parameters - supplied_keys))
raise RuntimeError("Parameters are incorrect")
###################################
# Set the RNGs and make it all deterministic
###################################
np.random.seed(p.seed)
random.seed(p.seed)
torch.manual_seed(p.seed)
torch.use_deterministic_algorithms(True)
###########################################
# The stratified datasets honor this
###########################################
torch.set_default_dtype(eval(p.torch_default_dtype))
###################################
# Build the network(s)
# Note: It's critical to do this AFTER setting the RNG
# (This is due to the randomized initial weights)
###################################
x_net = build_sequential(p.x_net)
start_time_secs = time.time()
###################################
# Build the dataset
###################################
if p.x_transforms_source == []: x_transform_source = None
else: x_transform_source = get_chained_transform(p.x_transforms_source)
if p.x_transforms_target == []: x_transform_target = None
else: x_transform_target = get_chained_transform(p.x_transforms_target)
if p.episode_transforms_source == []: episode_transform_source = None
else: raise Exception("episode_transform_source not implemented")
if p.episode_transforms_target == []: episode_transform_target = None
else: raise Exception("episode_transform_target not implemented")
eaf_source = Episodic_Accessor_Factory(
labels=p.labels_source,
domains=p.domains_source,
num_examples_per_domain_per_label=p.num_examples_per_domain_per_label_source,
iterator_seed=p.seed,
dataset_seed=p.dataset_seed,
n_shot=p.n_shot,
n_way=p.n_way,
n_query=p.n_query,
train_val_test_k_factors=(p.train_k_factor,p.val_k_factor,p.test_k_factor),
pickle_path=os.path.join(get_datasets_base_path(), p.pickle_name),
x_transform_func=x_transform_source,
example_transform_func=episode_transform_source,
)
train_original_source, val_original_source, test_original_source = eaf_source.get_train(), eaf_source.get_val(), eaf_source.get_test()
eaf_target = Episodic_Accessor_Factory(
labels=p.labels_target,
domains=p.domains_target,
num_examples_per_domain_per_label=p.num_examples_per_domain_per_label_target,
iterator_seed=p.seed,
dataset_seed=p.dataset_seed,
n_shot=p.n_shot,
n_way=p.n_way,
n_query=p.n_query,
train_val_test_k_factors=(p.train_k_factor,p.val_k_factor,p.test_k_factor),
pickle_path=os.path.join(get_datasets_base_path(), p.pickle_name),
x_transform_func=x_transform_target,
example_transform_func=episode_transform_target,
)
train_original_target, val_original_target, test_original_target = eaf_target.get_train(), eaf_target.get_val(), eaf_target.get_test()
transform_lambda = lambda ex: ex[1] # Original is (<domain>, <episode>) so we strip down to episode only
train_processed_source = Lazy_Iterable_Wrapper(train_original_source, transform_lambda)
val_processed_source = Lazy_Iterable_Wrapper(val_original_source, transform_lambda)
test_processed_source = Lazy_Iterable_Wrapper(test_original_source, transform_lambda)
train_processed_target = Lazy_Iterable_Wrapper(train_original_target, transform_lambda)
val_processed_target = Lazy_Iterable_Wrapper(val_original_target, transform_lambda)
test_processed_target = Lazy_Iterable_Wrapper(test_original_target, transform_lambda)
datasets = EasyDict({
"source": {
"original": {"train":train_original_source, "val":val_original_source, "test":test_original_source},
"processed": {"train":train_processed_source, "val":val_processed_source, "test":test_processed_source}
},
"target": {
"original": {"train":train_original_target, "val":val_original_target, "test":test_original_target},
"processed": {"train":train_processed_target, "val":val_processed_target, "test":test_processed_target}
},
})
# Some quick unit tests on the data
from steves_utils.transforms import get_average_power, get_average_magnitude
q_x, q_y, s_x, s_y, truth = next(iter(train_processed_source))
assert q_x.dtype == eval(p.torch_default_dtype)
assert s_x.dtype == eval(p.torch_default_dtype)
print("Visually inspect these to see if they line up with expected values given the transforms")
print('x_transforms_source', p.x_transforms_source)
print('x_transforms_target', p.x_transforms_target)
print("Average magnitude, source:", get_average_magnitude(q_x[0].numpy()))
print("Average power, source:", get_average_power(q_x[0].numpy()))
q_x, q_y, s_x, s_y, truth = next(iter(train_processed_target))
print("Average magnitude, target:", get_average_magnitude(q_x[0].numpy()))
print("Average power, target:", get_average_power(q_x[0].numpy()))
###################################
# Build the model
###################################
model = Steves_Prototypical_Network(x_net, device=p.device, x_shape=(2,256))
optimizer = Adam(params=model.parameters(), lr=p.lr)
###################################
# train
###################################
jig = PTN_Train_Eval_Test_Jig(model, p.BEST_MODEL_PATH, p.device)
jig.train(
train_iterable=datasets.source.processed.train,
source_val_iterable=datasets.source.processed.val,
target_val_iterable=datasets.target.processed.val,
num_epochs=p.n_epoch,
num_logs_per_epoch=p.NUM_LOGS_PER_EPOCH,
patience=p.patience,
optimizer=optimizer,
criteria_for_best=p.criteria_for_best,
)
total_experiment_time_secs = time.time() - start_time_secs
###################################
# Evaluate the model
###################################
source_test_label_accuracy, source_test_label_loss = jig.test(datasets.source.processed.test)
target_test_label_accuracy, target_test_label_loss = jig.test(datasets.target.processed.test)
source_val_label_accuracy, source_val_label_loss = jig.test(datasets.source.processed.val)
target_val_label_accuracy, target_val_label_loss = jig.test(datasets.target.processed.val)
history = jig.get_history()
total_epochs_trained = len(history["epoch_indices"])
val_dl = Iterable_Aggregator((datasets.source.original.val,datasets.target.original.val))
confusion = ptn_confusion_by_domain_over_dataloader(model, p.device, val_dl)
per_domain_accuracy = per_domain_accuracy_from_confusion(confusion)
# Add a key to per_domain_accuracy for if it was a source domain
for domain, accuracy in per_domain_accuracy.items():
per_domain_accuracy[domain] = {
"accuracy": accuracy,
"source?": domain in p.domains_source
}
# Do an independent accuracy assesment JUST TO BE SURE!
# _source_test_label_accuracy = independent_accuracy_assesment(model, datasets.source.processed.test, p.device)
# _target_test_label_accuracy = independent_accuracy_assesment(model, datasets.target.processed.test, p.device)
# _source_val_label_accuracy = independent_accuracy_assesment(model, datasets.source.processed.val, p.device)
# _target_val_label_accuracy = independent_accuracy_assesment(model, datasets.target.processed.val, p.device)
# assert(_source_test_label_accuracy == source_test_label_accuracy)
# assert(_target_test_label_accuracy == target_test_label_accuracy)
# assert(_source_val_label_accuracy == source_val_label_accuracy)
# assert(_target_val_label_accuracy == target_val_label_accuracy)
experiment = {
"experiment_name": p.experiment_name,
"parameters": dict(p),
"results": {
"source_test_label_accuracy": source_test_label_accuracy,
"source_test_label_loss": source_test_label_loss,
"target_test_label_accuracy": target_test_label_accuracy,
"target_test_label_loss": target_test_label_loss,
"source_val_label_accuracy": source_val_label_accuracy,
"source_val_label_loss": source_val_label_loss,
"target_val_label_accuracy": target_val_label_accuracy,
"target_val_label_loss": target_val_label_loss,
"total_epochs_trained": total_epochs_trained,
"total_experiment_time_secs": total_experiment_time_secs,
"confusion": confusion,
"per_domain_accuracy": per_domain_accuracy,
},
"history": history,
"dataset_metrics": get_dataset_metrics(datasets, "ptn"),
}
ax = get_loss_curve(experiment)
plt.show()
get_results_table(experiment)
get_domain_accuracies(experiment)
print("Source Test Label Accuracy:", experiment["results"]["source_test_label_accuracy"], "Target Test Label Accuracy:", experiment["results"]["target_test_label_accuracy"])
print("Source Val Label Accuracy:", experiment["results"]["source_val_label_accuracy"], "Target Val Label Accuracy:", experiment["results"]["target_val_label_accuracy"])
json.dumps(experiment)
###Output
_____no_output_____ |
MNIST_neuralnetworks.ipynb | ###Markdown
Intro to Pattern Recognition Problem Set 4: Neural Networks M S Mohamed Fazil UB Id : mm549 IntroductionWe will be using neural networks to build a model that classifies hand written digits using thr MNIST database. The MNIST database (Modified National Institute of Standards and Technology Database) is a large collection of handwritten digits which happens to be a state of the art data set to train and test machine learning models using image processing.It consist of 60000 Training samples and 10000 Testing Samples. Each sample image consist of normalized $28X28$ pixels grayscale image stored as 784 dimensional vector for each sample.This is a way of dimension reduction done by the MNIST to use it for model developments. Libraries UsedFor loading the given MNIST datasets we will be using the mnist library to fetch the data.The MNIST() function will be used to fetch the datasets from the directory './data'. For the purpose of building neural networks and trainning we will be using Keras library which is backed by the Tensorflow. We will use OpenCv for the image processing applications.
###Code
from keras.models import Sequential
from keras.layers import Dense, Activation, Conv2D, MaxPooling2D, Dropout, Flatten
from keras.callbacks import LambdaCallback,ModelCheckpoint
from keras.utils import to_categorical
from keras import optimizers,regularizers
import numpy as np
import cv2
import os
from mnist.loader import MNIST
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
import random as rd
os.environ["CUDA_VISIBLE_DEVICES"]="1"
%matplotlib inline
m = MNIST('./data')
###Output
Using TensorFlow backend.
###Markdown
1. Neural Network to Maximize the log Likehood Minimization of Negative Log Likehood with Softmax Output LayerSoftmax is an activation function of a node which is commonly placed as the output layer of a Neural Network. It is mostly used in multiclass classification problems. The Function is defined as \begin{equation}S(V_i) = \frac{e^{V_i}}{\sum_je^{V_i}}\\end{equation}The softmax function squashes the input vector values between 0 to 1 and because of the normalization of exponential , the sum of the values in the vector tends to be 1. The output values of a Softmax function can be interpretted as a probabilities of the different labels present in the multiclass.In our case the Softmax outputs a vector of 10 values corresponding to the class labels.Negative Log-Likehood is a loss function (aka criterion function) which is used with the Softmax activation function to minimize the criterion function. It is given as \begin{equation}J_0 = -log\,p(D;w) = -log\,p(\{(x_n,t_n):n=1,2,4,...\};w)\end{equation}\begin{equation}J_0 =-log\prod_n\prod_{m=0}^9p(t_n=m|x_n;w)\end{equation}To minimize the criterion function we must differentiate the softmax function with respect to the Negative Log Likehood function. Neural Network to Maximize the Posterior Likehood by Minimizing Criterion Function with L2 RegularizationLet us assume that a Gaussion Prior of the weight distribution is given with the $\alpha$.\begin{equation}p(w;\alpha)=N(N,\alpha l)\end{equation}Regularization is used to avoid overfitting of the traning datasets into the model.Therefore depending on the gradient of the weights between the epochs a part of the weights is subtracted by the regularization component determined with the regularization parameter $\alpha$. The Gaussian Prior function can be used in par with the L2 regularization during the minimization of the Criterion function which will maximize the Posterior Likehood. The combined loss function can be given as \begin{equation}J(w) = J_0(w)-log\,p(w;\alpha^{-1})\end{equation} Preparing the DatasetsWe prepare a dataset which will consist of 1000 training samples with labels where 100 images for each digit and another 1000 testing samples with label where 100 images for each digit. The target consist of a class with ten classification labels.
###Code
classes = [0,1,2,3,4,5,6,7,8,9]
xtrain,ytrain = m.load_training()
xtest,ytest = m.load_testing()
# Divide the given datasets to 1000 training dataset and 1000 testing dataset
xtrain = xtrain[0:1000]
ytrain = ytrain[0:1000]
xtest = xtest[0:1000]
ytest = ytest[0:1000]
xtrain = np.asarray(xtrain).astype(np.float32)
ytrain = np.asarray(ytrain).astype(np.float32)
xtest = np.asarray(xtest).astype(np.float32)
ytest = np.asarray(ytest).astype(np.float32)
n_classes = len(classes)
#0-1 Hot encoding
label_train = np.zeros((ytrain.shape[0], n_classes))
a = np.arange(ytrain.shape[0], dtype=np.int64)
b = np.array(ytrain, dtype=np.int64).reshape((ytrain.shape[0],))
label_train[a, b] = 1
label_test = np.zeros((ytest.shape[0], n_classes))
c = np.arange(ytest.shape[0], dtype=np.int64)
d = np.array(ytest, dtype=np.int64).reshape((ytest.shape[0],))
label_test[c, d] = 1
###Output
_____no_output_____
###Markdown
Function to Plot the Results of a Training ModelThis fuction plots the Traing and Testing's loss,criterion function, accuracy and the learning speed of each hidden layer with respect to each epoch during training phase. The learning Speed is the measure of the speed of change of weights after every epoch.
###Code
def plot_model_result(history,weights,final_w):
plt.plot(history.history['loss'], label='Training data')
plt.plot(history.history['val_loss'], label='Testing data')
plt.legend(['train', 'test'], loc='upper left')
plt.title('Loss of Training and Testing Data (Criterion Function)')
plt.ylabel('MAE value')
plt.xlabel('No. epoch')
plt.show()
plt.plot(history.history['accuracy'], label='Training data')
plt.plot(history.history['val_accuracy'], label='Testing data')
plt.legend(['train', 'test'], loc='upper left')
plt.title('Accuracy of Training and Testing Data')
plt.ylabel('Accuracy value')
plt.xlabel('No. epoch')
plt.show()
cnt = 0
'''
Plot the Learning Speed of the Model
'''
np.seterr(divide='ignore', invalid='ignore')
for i in range(0,len(weights)):
w = weights[i]
W = final_w[i]
learning_speed = [0]
for i in range(1,len(w)):
wold = w[i-1]
wnew = w[i]
diff = wold-wnew
d = wnew
m = np.median(d[d>0])
d[d==0] = m
diff = abs(diff)/d
if diff.size:
avg = abs(np.mean(diff))
else:
avg = 0
learning_speed.append(avg)
plt.plot(learning_speed, label='Learning Speed')
cnt = cnt+1
plt.title('Learning Speed of Hidden Layer {}'.format(cnt))
plt.ylabel('Learning Speed')
plt.xlabel('No. epoch')
plt.show()
print('Training Accuracy -',history.history['accuracy'][29])
print('Training Loss -',history.history['loss'][29])
print('Testing Accuracy -',history.history['val_accuracy'][29])
print('Testing Loss -',history.history['val_loss'][29])
###Output
_____no_output_____
###Markdown
2(a). Neural Network with 1 Hidden Layer without regularization
###Code
model = Sequential()
model.add(Dense(30, activation='sigmoid', input_dim=xtrain.shape[1]))
model.add(Dense(10, activation='softmax'))
weights = []
save_weights = LambdaCallback(on_epoch_end=lambda batch, logs: weights.append(model.layers[0].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(xtrain, label_train,validation_data=(xtest, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights])
final_w = [model.layers[0].get_weights()[0]]
weights = [weights]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____
###Markdown
2(b). Neural Network with 2 Hidden Layer and 3 Hidden Layer 2 Hidden Layer without Regularization
###Code
model = Sequential()
model.add(Dense(30, activation='sigmoid', input_dim=xtrain.shape[1]))
model.add(Dense(30, activation='sigmoid'))
model.add(Dense(10, activation='softmax'))
weights1 = []
weights2 = []
save_weights1 = LambdaCallback(on_epoch_end=lambda batch, logs: weights1.append(model.layers[0].get_weights()[0]))
save_weights2 = LambdaCallback(on_epoch_end=lambda batch, logs: weights2.append(model.layers[1].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(xtrain, label_train,validation_data=(xtest, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights1,save_weights2])
final_w = [model.layers[0].get_weights()[0],model.layers[1].get_weights()[0]]
weights = [weights1,weights2]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____
###Markdown
3 Hidden Layer Without Regularization
###Code
model = Sequential()
model.add(Dense(30, activation='sigmoid', input_dim=xtrain.shape[1]))
model.add(Dense(30, activation='sigmoid'))
model.add(Dense(30, activation='sigmoid'))
model.add(Dense(10, activation='softmax'))
weights1 = []
weights2 = []
weights3 = []
save_weights1 = LambdaCallback(on_epoch_end=lambda batch, logs: weights1.append(model.layers[0].get_weights()[0]))
save_weights2 = LambdaCallback(on_epoch_end=lambda batch, logs: weights2.append(model.layers[1].get_weights()[0]))
save_weights3 = LambdaCallback(on_epoch_end=lambda batch, logs: weights3.append(model.layers[1].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(xtrain, label_train,validation_data=(xtest, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights1,save_weights2,save_weights3])
final_w = [model.layers[0].get_weights()[0],model.layers[1].get_weights()[0],model.layers[2].get_weights()[0]]
weights = [weights1,weights2,weights3]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____
###Markdown
1 Hidden layer with Regularization - l2(5)
###Code
model = Sequential()
model.add(Dense(30, activation='sigmoid', input_dim=xtrain.shape[1],
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
model.add(Dense(10, activation='softmax',
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
weights = []
save_weights = LambdaCallback(on_epoch_end=lambda batch, logs: weights.append(model.layers[0].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(xtrain, label_train,validation_data=(xtest, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights])
final_w = [model.layers[0].get_weights()[0]]
weights = [weights]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____
###Markdown
2 Hidden Layer with Regularization - l2(5)
###Code
model = Sequential()
model.add(Dense(30, activation='sigmoid', input_dim=xtrain.shape[1],
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
model.add(Dense(30, activation='sigmoid',
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
model.add(Dense(10, activation='softmax',
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
weights1 = []
weights2 = []
save_weights1 = LambdaCallback(on_epoch_end=lambda batch, logs: weights1.append(model.layers[0].get_weights()[0]))
save_weights2 = LambdaCallback(on_epoch_end=lambda batch, logs: weights2.append(model.layers[1].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(xtrain, label_train,validation_data=(xtest, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights1,save_weights2])
final_w = [model.layers[0].get_weights()[0],model.layers[1].get_weights()[0]]
weights = [weights1,weights2]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____
###Markdown
3 Hidden Layer with Regularization - l2(5)
###Code
model = Sequential()
model.add(Dense(30, activation='sigmoid', input_dim=xtrain.shape[1],
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
model.add(Dense(30, activation='sigmoid',
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
model.add(Dense(30, activation='sigmoid',
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
model.add(Dense(10, activation='softmax',
kernel_regularizer=regularizers.l2(5),
bias_regularizer=regularizers.l2(5)))
weights1 = []
weights2 = []
weights3 = []
save_weights1 = LambdaCallback(on_epoch_end=lambda batch, logs: weights1.append(model.layers[0].get_weights()[0]))
save_weights2 = LambdaCallback(on_epoch_end=lambda batch, logs: weights2.append(model.layers[1].get_weights()[0]))
save_weights3 = LambdaCallback(on_epoch_end=lambda batch, logs: weights3.append(model.layers[1].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(xtrain, label_train,validation_data=(xtest, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights1,save_weights2,save_weights3])
final_w = [model.layers[0].get_weights()[0],model.layers[1].get_weights()[0],model.layers[2].get_weights()[0]]
weights = [weights1,weights2,weights3]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____
###Markdown
2(c) . Construct and Train Convolutional Neural networkFor Convolutional Neural Networks we first change shape of th Datasets from set of array of 1 Dimensional array of $28x28$ image to a set of 3-Dimensinal Image with Value.
###Code
x_train_imgs = xtrain.reshape(xtrain.shape[0],28,28,1)
x_test_imgs = xtrain.reshape(xtest.shape[0],28,28,1)
###Output
_____no_output_____
###Markdown
Preprocessing the Image :We use th ImageDataGenerator library from keras to generate a augumented form of out original training dataset.
###Code
from keras.preprocessing.image import ImageDataGenerator
aug_data = ImageDataGenerator(rotation_range=3,
width_shift_range=0.25, # 0.25 * 28 = 8 pixels approx
height_shift_range=0.25, # 0.25 * 28 = 8 pixels approx
horizontal_flip=True,
fill_mode="nearest")
aug_data.fit(x_train_imgs)
n = 0
for x_batch, y_batch in aug_data.flow(x_train_imgs, label_train, batch_size=1000):
x_batch = np.reshape(x_batch, [-1, 28*28])
model.fit(x_batch, y_batch)
n += 1
if n >= len(x_train_imgs) / 1000:
break
x_batch = x_batch.reshape(x_batch.shape[0],28,28,1)
###Output
Epoch 1/1
1000/1000 [==============================] - 0s 78us/step - loss: 0.3252 - accuracy: 0.9000
###Markdown
Now we build the Convolutional Neutral Network and feed the Augumented Training Dataset to our model.
###Code
model = Sequential()
model.add(Conv2D(30, activation='sigmoid',kernel_size=(5, 5), input_shape=(28,28,1)))
model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2)))
model.add(Flatten())
model.add(Dropout(0.25))
model.add(Dense(10, activation='softmax'))
weights = []
save_weights = LambdaCallback(on_epoch_end=lambda batch, logs: weights.append(model.layers[0].get_weights()[0]))
sgd = optimizers.SGD(learning_rate=0.1)
model.compile(optimizer=sgd,
loss='binary_crossentropy',
metrics=['accuracy'])
history = model.fit(x_batch, y_batch,validation_data=(x_test_imgs, label_test),
epochs=30, batch_size=10, shuffle = False,callbacks = [save_weights])
final_w = [model.layers[0].get_weights()[0]]
weights = [weights]
plot_model_result(history,weights,final_w)
###Output
_____no_output_____ |
notebooks/currents.ipynb | ###Markdown
Strengths of currentsThe notebook calculates the strength of the California Current and California Undercurrent for a domain similar to "A climatology of the California CurrentSystem from a network of underwater gliders" (Rudnick, 2017). The following figures are created in this notebook:- Figure A.9 California Current and Undercurrent strength
###Code
import sys
sys.path.append('/nfs/kryo/work/maxsimon/master-thesis/scripts/')
import os
import xarray as xr
import numpy as np
from matplotlib import pyplot as plt
plt.rcParams.update({'font.size': 12})
from romstools.romsrun import RomsRun
from romstools.plot import plot_data
from romstools.utils import p, cache, np_rolling_mean
from romstools.cmap import DIFF, DIFF_r, W2G, W2G_r, G2R, G2R_r, get_step_cmap
from romstools.dataset import open_dataset
import scipy.signal as sig
import scipy.stats as stat
import cartopy.crs as ccrs
import warnings
from datetime import timedelta as tdelta
import matplotlib.animation as animation
from matplotlib import rc
from matplotlib.cm import get_cmap
###Output
_____no_output_____
###Markdown
Load Data
###Code
### pactcs30
meso = RomsRun('/nfs/kryo/work/maxsimon/data/pactcs30/grid.nc')
# add location of zlevels
meso.add_data('/nfs/kryo/work/maxsimon/data/pactcs30/z/z_levels.nc')
# add horizontal velocities
meso.add_data('/nfs/kryo/work/maxsimon/data/pactcs30/climatologies/z_vel-1d.nc')
# add density
meso.add_data('/nfs/kryo/work/maxsimon/data/pactcs30/climatologies/z_data-1d.nc')
# add additional grid data
data = np.load('/nfs/kryo/work/maxsimon/data/pactcs30/grid.npz', allow_pickle=True)
meso.distance_map = data['distance_map']
meso.distance_lines = data['distance_lines']
meso.gruber_mask = data['gruber_mask']
### pactcs15
subm = RomsRun('/nfs/kryo/work/maxsimon/data/pactcs15/grid.nc')
# add location of zlevels
subm.add_data('/nfs/kryo/work/maxsimon/data/pactcs15/z/z_levels.nc')
# add horizontal velocities
subm.add_data('/nfs/kryo/work/maxsimon/data/pactcs15/climatologies/z_vel-1d.nc')
# add density
subm.add_data('/nfs/kryo/work/maxsimon/data/pactcs15/climatologies/z_data-1d.nc')
# add additional grid data
data = np.load('/nfs/kryo/work/maxsimon/data/pactcs15/grid.npz', allow_pickle=True)
subm.distance_map = data['distance_map']
subm.distance_lines = data['distance_lines']
subm.gruber_mask = data['gruber_mask']
runs = {
'pactcs15': subm,
'pactcs30': meso
}
###Output
_____no_output_____
###Markdown
SubdomainThe subdomain is defined such that it corresponds to> Rudnick, Daniel L. et al. (May 2017). “A climatology of the California CurrentSystem from a network of underwater gliders.”, Figure 4.2.3.1
###Code
SLICE_SUBM_1 = (slice(None, None), slice(640, 720))
SLICE_MESO_1 = (slice(None, None), slice(261, 296))
vertical_sections = {
'pactcs30': SLICE_MESO_1,
'pactcs15': SLICE_SUBM_1
}
# Plot subdomain
def plot_run_slices(run, ax, slices):
for s in slices:
plot_data(run.grid, run.distance_map, ax=ax, highlight_subdomain=s, lon_name='lon_rho', lat_name='lat_rho', as_contourfill=True, cbar_label='Distance [km]', cmap='Blues', highlight_subdomain_alpha=0.0, colorbar=False);
for line in run.distance_lines:
ax.plot(line[0], line[1], color='white', transform=ccrs.PlateCarree())
ax.set_ylim(24, 42)
ax.set_xlim(-136, -120)
fig, ax = plt.subplots(1, 2, figsize=(20, 10), subplot_kw={'projection': ccrs.PlateCarree()})
plot_run_slices(subm, ax[0], [SLICE_SUBM_1])
plot_run_slices(meso, ax[1], [SLICE_MESO_1])
plt.show()
###Output
/nfs/kryo/work/maxsimon/master-thesis/scripts/romstools/plot.py:123: RuntimeWarning: invalid value encountered in greater
data_main[data_main > vmax] = vmax
/nfs/kryo/work/maxsimon/master-thesis/scripts/romstools/plot.py:124: RuntimeWarning: invalid value encountered in less
data_main[data_main < vmin] = vmin
/nfs/kryo/work/maxsimon/master-thesis/scripts/romstools/plot.py:128: RuntimeWarning: invalid value encountered in greater
data_subd[data_subd > vmax] = vmax
/nfs/kryo/work/maxsimon/master-thesis/scripts/romstools/plot.py:129: RuntimeWarning: invalid value encountered in less
data_subd[data_subd < vmin] = vmin
###Markdown
Plotting and interpolation
###Code
# see eddy_quenching for an analysis of this
interp_points = {
'pactcs15': 200,
'pactcs30': 100
}
## This is a copy from eddy_quenching.
## TODO: move to external module
def fix_nan_contour(func, distance_map, depth_values, data, **kwargs):
# matplotlibs contour plots can not handle NaNs
# this function fixes this
xx, yy = None, None
if len(distance_map.shape) == 2:
print('WARNING :: DISTANCE MAP ON SECTION SLICE?')
contour_x = np.mean(distance_map, axis=1)
x_1 = np.argmin(np.isnan(contour_x[::-1]))
xx, yy = np.meshgrid(contour_x[:-x_1], -depth_values)
else:
contour_x = distance_map
xx, yy = np.meshgrid(contour_x, -depth_values)
return func(xx[:data.shape[0]], yy[:data.shape[0]], data, **kwargs)
def interpolate_to_dist(data, name, num_interp_points, distance_map=None):
# interpolate data on rho grid to a grid with distance to coast as the main axis.
# get run and distance map
run = runs[name]
dmap = run.distance_map[vertical_sections[name]] if distance_map is None else distance_map
# set up bins
distances = np.linspace(0, 900, num_interp_points)
# and result array
result = np.empty((data.shape[0], distances.shape[0] - 1))
centers = []
# loop bins
for dist_idx in range(distances.shape[0] - 1):
# create mask
mask = np.logical_and(
dmap >= distances[dist_idx],
dmap < distances[dist_idx + 1]
)
# calculate the value as average over all points belonging to the bin
value = np.nanmean(
data[:, mask], axis=1
)
# assign value to result
result[:, dist_idx] = value
# save the bin center for x coordinates
centers.append(distances[dist_idx] + (distances[dist_idx + 1] - distances[dist_idx])/2)
return np.array(centers), result
def plot_vertical(self, ax, distances, values, vmin=None, vmax=None, num_levels=30, num_levels_lines=10, cmap=None, colorbar_label='', contour_lines=None, colorbar=True):
# set limits
ax.set_xlim(900, 0)
# use number of levels or contstruct levels from vmin and vmax
levels = num_levels if vmin is None or vmax is None else np.linspace(vmin, vmax, num_levels)
# plot data
cax = fix_nan_contour(ax.contourf, distances, self.z_level, values, levels=levels, vmin=vmin, vmax=vmax, cmap=cmap, extend='both')
# get data for contours
contour_values = contour_lines if contour_lines is not None else values
# plot contours
if num_levels_lines > 0:
cax2 = fix_nan_contour(ax.contour, distances, self.z_level, contour_values, levels=num_levels_lines, colors='k', extend='both')
ax.clabel(cax2, cax2.levels, inline=True, fontsize=10)
# labels
ax.set_xlabel('Distance to coast [km]')
ax.set_ylabel('Depth [m]')
# colorbar
ticks = None # if vmin and vmax is provided, set up ticks for colorbar manually
if vmin is not None and vmax is not None:
ticks = np.linspace(vmin, vmax, 11)
if colorbar:
plt.colorbar(cax, ticks=ticks, label=colorbar_label, ax=ax)
return cax
###Output
_____no_output_____
###Markdown
Calculate Velocity
###Code
def calc_u_vertical(self, section_slice, time_slice, name, var_contour='u_b'):
"""
Calculate a vertical section of u (parrallel to coast) for a given xi-section and time slice
"""
num_interp_points = interp_points[name]
# calculate u values in cm/s
u_b_values = self['u_b'].isel(xi_rho=section_slice[1], doy=time_slice).mean(dim=['doy']).values * 100
# calculate contour values
values_contour = None if var_contour == 'u_b' else self[var_contour].isel(xi_rho=section_slice[1], doy=time_slice).mean(dim=['doy']).values
# interpolate data to distance to coast space
u_b_interpolated = interpolate_to_dist(u_b_values, name, num_interp_points, distance_map=self.distance_map[section_slice])
# u_b_interpolated = interpolate_to_dist(u_b_values, self.distance_map[section_slice], distances=np.linspace(0, 900, num_interp_points))
# interpolate contours to distance from coast space
# contour_interpolated = None if var_contour == 'u_b' else interpolate_to_dist(values_contour, self.distance_map[section_slice], distances=np.linspace(0, 900, num_interp_points))
contour_interpolated = None if var_contour == 'u_b' else interpolate_to_dist(values_contour, name, num_interp_points, distance_map=self.distance_map[section_slice])
return u_b_interpolated, contour_interpolated
def calc_vertical_comparison(slices, tslice):
"""
Calculate the vertical section for both, pactcs15 and pactcs30
"""
res = {}
for i, slice in enumerate(slices):
res[i] = {
'subm': calc_u_vertical(subm, slice[0], tslice, 'pactcs15', var_contour='rho_b'),
'meso': calc_u_vertical(meso, slice[1], tslice, 'pactcs30', var_contour='rho_b'),
}
return res
# set up doys
all_year = np.arange(365)
t0 = all_year[30:90] # february and march
t1 = all_year[180:270] # june to august
# calculate u-component for t0
res_t0 = calc_vertical_comparison([
(SLICE_SUBM_1, SLICE_MESO_1)
], t0)
# calculate u-component for t1
res_t1 = calc_vertical_comparison([
(SLICE_SUBM_1, SLICE_MESO_1)
], t1)
# join to dictionary
res = {
0: res_t0[0],
1: res_t1[0]
}
def plot_precalculated_comparison(res, path='', captions=[]):
fig, ax = plt.subplots(len(res), 2, figsize=(15, 5*len(res)), sharex=True, sharey=True)
# lets us use indexing on ax even for a single item
if len(res) == 1:
ax = np.array([ax])
# loop the results to compare
for i in range(len(res)):
add_caption = '' if len(captions) == 0 else ' - '+captions[i]
# get values
f_b_interpolated, contour_interpolated = res[i]['meso']
# plot
plot_vertical(meso, ax[i, 0], f_b_interpolated[0], f_b_interpolated[1], vmin=-25, vmax=25, cmap='bwr_r', contour_lines=contour_interpolated[1], colorbar=False)
# add title and extent
ax[i, 0].set_title('MR' + add_caption)
ax[i, 0].set_xlim(600, 0)
# get values
f_b_interpolated, contour_interpolated = res[i]['subm']
# plot
cax = plot_vertical(subm, ax[i, 1], f_b_interpolated[0], f_b_interpolated[1], vmin=-25, vmax=25, cmap='bwr_r', contour_lines=contour_interpolated[1], colorbar=False)
# add title and extent
ax[i, 1].set_title('HR' + add_caption)
ax[i, 1].set_xlim(600, 0)
# set xlim and ylim
ax[i, 0].set_xlim(250, 0)
ax[i, 1].set_xlim(250, 0)
ax[i, 0].set_ylim(-500, 0)
ax[i, 1].set_ylim(-500, 0)
# add or remove labels
ax[i, 1].set_ylabel('')
if i != len(res) - 1:
ax[i, 0].set_xlabel('')
ax[i, 1].set_xlabel('')
# colorbar
colorbar_label = 'North u [cm / s] South'
plt.colorbar(cax, ax=ax, label=colorbar_label, location='bottom', ticks=[-25, -20, -15, -10, -5, 0, 5, 10, 15, 20, 25])
if path != '':
plt.savefig(path)
plt.show()
plot_precalculated_comparison(res, captions=['February to March', 'June to August'], path='figures/result_undercurrent.pdf')
###Output
_____no_output_____ |
spot/tests/python/ltsmin-dve.ipynb | ###Markdown
There are two ways to load a DiVinE model: from a file or from a cell. Loading from a file-------------------We will first start with the file version, however because this notebook should also be a self-contained test case, we start by writing a model into a file.
###Code
!rm -f test1.dve
%%file test1.dve
int a = 0, b = 0;
process P {
state x;
init x;
trans
x -> x { guard a < 3 && b < 3; effect a = a + 1; },
x -> x { guard a < 3 && b < 3; effect b = b + 1; };
}
process Q {
state wait, work;
init wait;
trans
wait -> work { guard b > 1; },
work -> wait { guard a > 1; };
}
system async;
###Output
Writing test1.dve
###Markdown
The `spot.ltsmin.load` function compiles the model using the `ltlmin` interface and load it. This should work with DiVinE models if `divine --LTSmin` works, and with Promela models if `spins` is installed.
###Code
m = spot.ltsmin.load('test1.dve')
###Output
_____no_output_____
###Markdown
Compiling the model creates all several kinds of files. The `test1.dve` file is converted into a C++ source code `test1.dve.cpp` which is then compiled into a shared library `test1.dve2c`. Becauce `spot.ltsmin.load()` has already loaded this shared library, all those files can be erased. If you do not erase the files, `spot.ltsmin.load()` will use the timestamps to decide whether the library should be recompiled or not everytime you load the library.For editing and loading DVE file from a notebook, it is a better to use the `%%dve` as shown next.
###Code
!rm -f test1.dve test1.dve.cpp test1.dve2C
###Output
_____no_output_____
###Markdown
Loading from a notebook cell----------------------------The `%%dve` cell magic implements all of the above steps (saving the model into a temporary file, compiling it, loading it, erasing the temporary files). The variable name that should receive the model (here `m`) should be indicated on the first line, after `%dve`.
###Code
%%dve m
int a = 0, b = 0;
process P {
state x;
init x;
trans
x -> x { guard a < 3 && b < 3; effect a = a + 1; },
x -> x { guard a < 3 && b < 3; effect b = b + 1; };
}
process Q {
state wait, work;
init wait;
trans
wait -> work { guard b > 1; },
work -> wait { guard a > 1; };
}
system async;
###Output
_____no_output_____
###Markdown
Working with an ltsmin model----------------------------Printing an ltsmin model shows some information about the variables it contains and their types, however the `info()` methods provide the data in a map that is easier to work with.
###Code
m
sorted(m.info().items())
###Output
_____no_output_____
###Markdown
To obtain a Kripke structure, call `kripke` and supply a list of atomic propositions to observe in the model.
###Code
k = m.kripke(["a<1", "b>2"])
k
k.show('.<15')
k.show('.<0') # unlimited output
a = spot.translate('"a<1" U "b>2"'); a
spot.otf_product(k, a)
###Output
_____no_output_____
###Markdown
If we want to create a `model_check` function that takes a model and formula, we need to get the list of atomic propositions used in the formula using `atomic_prop_collect()`. This returns an `atomic_prop_set`:
###Code
a = spot.atomic_prop_collect(spot.formula('"a < 2" W "b == 1"')); a
def model_check(f, m):
f = spot.formula(f)
ss = m.kripke(spot.atomic_prop_collect(f))
nf = spot.formula_Not(f).translate()
return spot.otf_product(ss, nf).is_empty()
model_check('"a<1" R "b > 1"', m)
###Output
_____no_output_____
###Markdown
Instead of `otf_product(x, y).is_empty()` we prefer to call `!x.intersects(y)`. There is also `x.intersecting_run(y)` that can be used to return a counterexample.
###Code
def model_debug(f, m):
f = spot.formula(f)
ss = m.kripke(spot.atomic_prop_collect(f))
nf = spot.formula_Not(f).translate()
return ss.intersecting_run(nf)
run = model_debug('"a<1" R "b > 1"', m); run
###Output
_____no_output_____
###Markdown
This accepting run can be represented as an automaton (the `True` argument requires the state names to be preserved). This can be more readable.
###Code
run.as_twa(True)
###Output
_____no_output_____ |
soln/glucose_soln.ipynb | ###Markdown
Modeling and Simulation in PythonChapter 18Copyright 2017 Allen DowneyLicense: [Creative Commons Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0)
###Code
# Configure Jupyter so figures appear in the notebook
%matplotlib inline
# Configure Jupyter to display the assigned value after an assignment
%config InteractiveShell.ast_node_interactivity='last_expr_or_assign'
# import functions from the modsim.py module
from modsim import *
###Output
_____no_output_____
###Markdown
Code from the previous chapterRead the data.
###Code
data = pd.read_csv('data/glucose_insulin.csv', index_col='time');
###Output
_____no_output_____
###Markdown
Interpolate the insulin data.
###Code
I = interpolate(data.insulin)
###Output
_____no_output_____
###Markdown
Initialize the parameters
###Code
G0 = 290
k1 = 0.03
k2 = 0.02
k3 = 1e-05
###Output
_____no_output_____
###Markdown
To estimate basal levels, we'll use the concentrations at `t=0`.
###Code
Gb = data.glucose[0]
Ib = data.insulin[0]
###Output
_____no_output_____
###Markdown
Create the initial condtions.
###Code
init = State(G=G0, X=0)
###Output
_____no_output_____
###Markdown
Make the `System` object.
###Code
t_0 = get_first_label(data)
t_end = get_last_label(data)
system = System(G0=G0, k1=k1, k2=k2, k3=k3,
init=init, Gb=Gb, Ib=Ib, I=I,
t_0=t_0, t_end=t_end, dt=2)
def update_func(state, t, system):
"""Updates the glucose minimal model.
state: State object
t: time in min
system: System object
returns: State object
"""
G, X = state
k1, k2, k3 = system.k1, system.k2, system.k3
I, Ib, Gb = system.I, system.Ib, system.Gb
dt = system.dt
dGdt = -k1 * (G - Gb) - X*G
dXdt = k3 * (I(t) - Ib) - k2 * X
G += dGdt * dt
X += dXdt * dt
return State(G=G, X=X)
def run_simulation(system, update_func):
"""Runs a simulation of the system.
system: System object
update_func: function that updates state
returns: TimeFrame
"""
t_0, t_end, dt = system.t_0, system.t_end, system.dt
frame = TimeFrame(columns=init.index)
frame.row[t_0] = init
ts = linrange(t_0, t_end, dt)
for t in ts:
frame.row[t+dt] = update_func(frame.row[t], t, system)
return frame
results = run_simulation(system, update_func);
###Output
_____no_output_____
###Markdown
Numerical solutionIn the previous chapter, we approximated the differential equations with difference equations, and solved them using `run_simulation`.In this chapter, we solve the differential equation numerically using `run_ode_solver`, which is a wrapper for the SciPy ODE solver.Instead of an update function, we provide a slope function that evaluates the right-hand side of the differential equations. We don't have to do the update part; the solver does it for us.
###Code
def slope_func(state, t, system):
"""Computes derivatives of the glucose minimal model.
state: State object
t: time in min
system: System object
returns: derivatives of G and X
"""
G, X = state
k1, k2, k3 = system.k1, system.k2, system.k3
I, Ib, Gb = system.I, system.Ib, system.Gb
dGdt = -k1 * (G - Gb) - X*G
dXdt = k3 * (I(t) - Ib) - k2 * X
return dGdt, dXdt
###Output
_____no_output_____
###Markdown
We can test the slope function with the initial conditions.
###Code
slope_func(init, 0, system)
###Output
_____no_output_____
###Markdown
Here's how we run the ODE solver.
###Code
results2, details = run_ode_solver(system, slope_func, t_eval=data.index);
###Output
_____no_output_____
###Markdown
`details` is a `ModSimSeries` object with information about how the solver worked.
###Code
details
###Output
_____no_output_____
###Markdown
`results` is a `TimeFrame` with one row for each time step and one column for each state variable:
###Code
results2
###Output
_____no_output_____
###Markdown
Plotting the results from `run_simulation` and `run_ode_solver`, we can see that they are not very different.
###Code
plot(results.G, '-')
plot(results2.G, '-')
plot(data.glucose, 'bo')
###Output
_____no_output_____
###Markdown
The differences in `G` are less than 1%.
###Code
diff = results.G - results2.G
percent_diff = diff / results2.G * 100
percent_diff.dropna()
###Output
_____no_output_____
###Markdown
Optimization Now let's find the parameters that yield the best fit for the data. We'll use these values as an initial estimate and iteratively improve them.
###Code
params = Params(G0 = 290,
k1 = 0.03,
k2 = 0.02,
k3 = 1e-05)
###Output
_____no_output_____
###Markdown
`make_system` takes the parameters and actual data and returns a `System` object.
###Code
def make_system(params, data):
"""Makes a System object with the given parameters.
params: sequence of G0, k1, k2, k3
data: DataFrame with `glucose` and `insulin`
returns: System object
"""
# params might be a Params object or an array,
# so we have to unpack it like this
G0, k1, k2, k3 = params
Gb = data.glucose[0]
Ib = data.insulin[0]
I = interpolate(data.insulin)
t_0 = get_first_label(data)
t_end = get_last_label(data)
init = State(G=G0, X=0)
return System(G0=G0, k1=k1, k2=k2, k3=k3,
init=init, Gb=Gb, Ib=Ib, I=I,
t_0=t_0, t_end=t_end, dt=2)
system = make_system(params, data)
###Output
_____no_output_____
###Markdown
`error_func` takes the parameters and actual data, makes a `System` object, and runs `odeint`, then compares the results to the data. It returns an array of errors.
###Code
system = make_system(params, data)
results, details = run_ode_solver(system, slope_func, t_eval=data.index)
details
def error_func(params, data):
"""Computes an array of errors to be minimized.
params: sequence of parameters
data: DataFrame of values to be matched
returns: array of errors
"""
print(params)
# make a System with the given parameters
system = make_system(params, data)
# solve the ODE
results, details = run_ode_solver(system, slope_func, t_eval=data.index)
# compute the difference between the model
# results and actual data
errors = results.G - data.glucose
return errors
###Output
_____no_output_____
###Markdown
When we call `error_func`, we provide a sequence of parameters as a single object. Here's how that works:
###Code
error_func(params, data)
###Output
G0 290.00000
k1 0.03000
k2 0.02000
k3 0.00001
dtype: float64
###Markdown
`leastsq` is a wrapper for `scipy.optimize.leastsq` Here's how we call it.
###Code
best_params, fit_details = leastsq(error_func, params, data)
###Output
[2.9e+02 3.0e-02 2.0e-02 1.0e-05]
[2.9e+02 3.0e-02 2.0e-02 1.0e-05]
[2.9e+02 3.0e-02 2.0e-02 1.0e-05]
[2.90000004e+02 3.00000000e-02 2.00000000e-02 1.00000000e-05]
[2.90000000e+02 3.00000004e-02 2.00000000e-02 1.00000000e-05]
[2.90000000e+02 3.00000000e-02 2.00000003e-02 1.00000000e-05]
[2.90000000e+02 3.00000000e-02 2.00000000e-02 1.00000001e-05]
###Markdown
The first return value is a `Params` object with the best parameters:
###Code
best_params
###Output
_____no_output_____
###Markdown
The second return value is a `ModSimSeries` object with information about the results.
###Code
fit_details
###Output
_____no_output_____
###Markdown
Now that we have `best_params`, we can use it to make a `System` object and run it.
###Code
system = make_system(best_params, data)
results, details = run_ode_solver(system, slope_func, t_eval=data.index)
details.message
###Output
_____no_output_____
###Markdown
Here are the results, along with the data. The first few points of the model don't fit the data, but we don't expect them to.
###Code
plot(results.G, label='simulation')
plot(data.glucose, 'bo', label='glucose data')
decorate(xlabel='Time (min)',
ylabel='Concentration (mg/dL)')
savefig('figs/chap08-fig04.pdf')
###Output
Saving figure to file figs/chap08-fig04.pdf
###Markdown
Interpreting parametersBased on the parameters of the model, we can estimate glucose effectiveness and insulin sensitivity.
###Code
def indices(params):
"""Compute glucose effectiveness and insulin sensitivity.
params: sequence of G0, k1, k2, k3
data: DataFrame with `glucose` and `insulin`
returns: State object containing S_G and S_I
"""
G0, k1, k2, k3 = params
return State(S_G=k1, S_I=k3/k2)
###Output
_____no_output_____
###Markdown
Here are the results.
###Code
indices(best_params)
###Output
_____no_output_____
###Markdown
Under the hoodHere's the source code for `run_ode_solver` and `leastsq`, if you'd like to know how they work.
###Code
source_code(run_ode_solver)
source_code(leastsq)
###Output
def leastsq(error_func, x0, *args, **options):
"""Find the parameters that yield the best fit for the data.
`x0` can be a sequence, array, Series, or Params
Positional arguments are passed along to `error_func`.
Keyword arguments are passed to `scipy.optimize.leastsq`
error_func: function that computes a sequence of errors
x0: initial guess for the best parameters
args: passed to error_func
options: passed to leastsq
:returns: Params object with best_params and ModSimSeries with details
"""
# override `full_output` so we get a message if something goes wrong
options['full_output'] = True
# run leastsq
t = scipy.optimize.leastsq(error_func, x0=x0, args=args, **options)
best_params, cov_x, infodict, mesg, ier = t
# pack the results into a ModSimSeries object
details = ModSimSeries(infodict)
details.set(cov_x=cov_x, mesg=mesg, ier=ier)
# if we got a Params object, we should return a Params object
if isinstance(x0, Params):
best_params = Params(Series(best_params, x0.index))
# return the best parameters and details
return best_params, details
|
lecture03-python-introduction/lecture03-with-solutions.ipynb | ###Markdown
Review exercise 1 menti.com 52 00 21 Review exercise 2 menti.com 74 95 17 Review exercise 3
###Code
def squared(x):
return x**2
###Output
_____no_output_____
###Markdown
Write a function ```pythonderivative(x, h, f)```which calculates the first numerical derivative of an arbitrary function $f(x)$. Use the formula $\frac{f(x + h) -f(x - h)}{2 h}$.Note: in Python you can pass a function as an argument of another function!Test it using parameters $x = 4$, $h = 0.0001$ and the function ```squared```defined above.
###Code
def derivative(x, h, f):
return (f(x+h) - f(x - h)) / (2 * h)
derivative(4, 0.0001, squared)
###Output
_____no_output_____
###Markdown
Lecture 3 - continuing the introduction to Python Data types in Python
###Code
type(1)
type(1.1)
type(1 + 2j)
type(True)
type("Hello World")
a = 1
type(a)
b = 1.1
type(b)
c = a + b
###Output
_____no_output_____
###Markdown
What is the type of c?
###Code
type(c)
d = 2
type(d)
e = a + d
###Output
_____no_output_____
###Markdown
What is the type of e?
###Code
type(e)
f = a / d
###Output
_____no_output_____
###Markdown
_Warning: Python 2 behaves differently here, but Python 2 is not relevant anylonger (unsupported since 1.1.2020)._ What is the type of f?
###Code
type(f)
###Output
_____no_output_____
###Markdown
Hints Integer overflowsInteger variables cannot overflow in Python, i.e. you can have an arbitrarily high Integer number. This is different to other programming languages such as C, where the size of an Integer number is fixed. If you do not know, what any of this means, it will be explained later when talking about numpy as numpy integers can overflow. Floating point precisionFloating point numbers have limited precision (53 bits or 16 digits). A number can become arbitrarily high, but it has limited precisions. This is due to how internally, floating points are stored in memory. More details on [wikipedia](https://en.wikipedia.org/wiki/Floating-point_arithmetic) and a tutorial can be found [here](http://cstl-csm.semo.edu/xzhang/Class%20Folder/CS280/Workbook_HTML/FLOATING_tut.htm)Attribution: Joeleoj123 / CC BY-SA (https://creativecommons.org/licenses/by-sa/4.0) Why is this relevant?
###Code
a_large_number = 10**12 + 0.1
a_large_number
a_small_number = 0.0000001
many_times = 10000000
for i in range(0, many_times):
a_large_number = a_large_number + a_small_number
a_large_number
a_large_number = a_large_number + many_times * a_small_number
a_large_number
###Output
_____no_output_____
###Markdown
Whaaaaaaaaat?Well - limited precisions.```pythona_large_number + a_small_number```is, due to limited precision, calculated to be```pythona_large_number + a_small_number``` String operations String is text. A new string is simply started by enclosing text in "
###Code
'This is a string'
###Output
_____no_output_____
###Markdown
Double quotes " and single quotes ' are equivalent Python, there is no syntactic difference. [A very popular convention](https://stackoverflow.com/a/56190/859591) is to use single quotes for constants and double quotes for human readable text or sentances. A different popular convention is to use only [only double quotes](https://black.readthedocs.io/en/stable/the_black_code_style.htmlstrings).
###Code
x = 7
f'The value of x is {x}'
import math
y = 6.589309493
f'The value of y is {round(y)}'
a_lot_of_text = (
'line1'
'line2'
'line3')
a_lot_of_text
###Output
_____no_output_____
###Markdown
But...
###Code
a_lot_of_text = """
line1
line2
line3"""
a_lot_of_text
###Output
_____no_output_____
###Markdown
Why is there a \n? Exercise 1Write code which shows the following text 'PI is '. Substitute by the value of PI. Hint: ```Pythonmath.pi```returns pi
###Code
import math
f"PI is {math.pi}"
###Output
_____no_output_____
###Markdown
Multiple assignments and return values
###Code
a, b = 1, 2
a
b
###Output
_____no_output_____
###Markdown
Functions with multiple return values
###Code
import math
def calculate_growth_doubling_time(value_t1, value_t2, time_diff):
growth_rate = (value_t2 / value_t1)**(1 / time_diff) - 1
doubling_time = math.log(2) / math.log(growth_rate + 1)
return growth_rate, doubling_time
growth_rate, doubling_time = calculate_growth_doubling_time(1, 1.2, 1)
print(growth_rate)
print(doubling_time)
###Output
0.19999999999999996
3.8017840169239308
###Markdown
Exercise 2Write a function that calculates the growth_rate, the doubling time, and the relative growth_rate given by $\frac{\textrm{value_t2} - \textrm{value_t1}}{\textrm{value_t1}}$. Use the upper function calculate_growth_doubling_time for that purpose. Return all three values. The function should be called ```pythongrowth_parameters(value_t1, value_t2, time_diff)```
###Code
def growth_parameters(value_t1, value_t2, time_diff):
return (value_t2 - value_t1) / value_t1
growth_parameters(1, 1.2, 1)
###Output
_____no_output_____
###Markdown
Some basic data structures in Python Lists
###Code
l = [1, 2, 3]
l
l[1]
###Output
_____no_output_____
###Markdown
Important: different to R, an index in Python starts with 0, not with 1. So the first element of a list is 0!
###Code
l[0]
###Output
_____no_output_____
###Markdown
What will happen here?
###Code
l[3]
l[0:2]
l = [1, 2, 3, 4, 5, 6]
l[1:5]
len(l)
###Output
_____no_output_____
###Markdown
Exercise 3Write a function named ```pythonlist_info(l)```that takes a list and returns the length of the list, the first element and the last element. Test the code with the list```python[7, 14, 21, 43]```
###Code
def list_info(l):
return len(l), l[0], l[-1]
list_info([7, 14, 21, 43])
###Output
_____no_output_____
###Markdown
Operators on lists
###Code
[1, 2] * 2
[1, 2] + [2]
[1, 2] + 2
###Output
_____no_output_____
###Markdown
Control structures: loopsLoops allow to repeat things. We will see that they become obsolete in many cases, when working with the scientific packages in Python (such as numpy), but we still introduce them.
###Code
for x in range(3):
print(x)
for x in range(17, 19):
print(x)
###Output
17
18
###Markdown
What will happen here?
###Code
for x in range(19, 17):
print(x)
l = [1, 2, 3]
for v in l:
print(v)
# other programing languages do it like this, in Python you never need this!
for k in range(0, len(l)):
print(l[k])
###Output
1
2
3
###Markdown
Pro: this may become important - if you do not understand it now, don't worry.Lists are passed by reference - whatever you do to a list in function, will also affect the list outside of the function. Watch!
###Code
def change_list(l):
l1 = l
l1[0] = 43
return l1
l = [1, 2, 3]
result = change_list(l)
print(l)
print(result)
def do_not_change_list(l):
l1 = l.copy()
l1[0] = 43
return l1
l = [1, 2, 3]
result = do_not_change_list(l)
print(l)
print(result)
###Output
[1, 2, 3]
[43, 2, 3]
###Markdown
List comprehensionsTo work with lists and for loops is not very Pythonic. In many cases list comprehensions are much more convenient:
###Code
l = [1, 10, 100]
# Take log10 of all list elements:
[math.log(i, 10) for i in l]
###Output
_____no_output_____
###Markdown
Exercise 4Write a function ```pythonlog10_list(l)```which calculates the logarithmus with basis 10 of all elements in that list and returns the resulting list. Test it on the list```pythonl = [1, 10, 100, 1000]```Print ```pythonl```before and after calling your function.
###Code
def log10_list(l):
return [math.log(i, 10) for i in l]
l = [1, 10, 100, 1000]
print("Before:", l)
result = log10_list(l)
print("After:", l)
print("Result: ", result)
###Output
Before: [1, 10, 100, 1000]
After: [1, 10, 100, 1000]
Result: [0.0, 1.0, 2.0, 2.9999999999999996]
###Markdown
Reading dataWe will read data from disk. This is just to make things more interesting. Do not worry about the details now.
###Code
import csv
with open('austria_covid_19_data.csv', 'r') as read_obj:
# pass the file object to reader() to get the reader object
csv_reader = csv.reader(read_obj)
# Pass reader object to list() to get a list of lists
list_of_rows = list(csv_reader)
list_of_rows
### get second column
infected_cases = [i[1] for i in list_of_rows]
infected_cases
### remove first line, as it contains header
infected_cases = infected_cases[1:]
infected_cases
### convert to integer, as it is string
infected_cases = [int(i) for i in infected_cases]
infected_cases
###Output
_____no_output_____
###Markdown
Plotting data with matplotlib
###Code
import math
import matplotlib.pyplot as plt
plt.plot(infected_cases, label = 'infections')
plt.plot([math.log(i, 10) for i in infected_cases], label = 'log(infections)')
plt.plot([infected_cases[0] * 1.28**i for i in range(len(infected_cases))], label = 'exponential with growth rate 28%')
plt.xlabel('Time')
plt.ylabel('Infections')
plt.title('Infections')
plt.legend()
plt.show()
import matplotlib.pyplot as plt
import math
plt.plot([math.log(i, 10) for i in infected_cases], label = 'log(infections)')
plt.xlabel('Time')
plt.ylabel('log(infections)')
plt.title('Log(infections)')
plt.legend()
plt.show()
###Output
_____no_output_____
###Markdown
Exercise 5Calculate the daily growth rate for all time steps and the doubling time for all time steps and plot both in one extra figure, i.e. one figure showing the growth rate, the other one showing the doubling time. For that purpose you can use the function```calculate_growth_doubling_time(value_t1, value_t2, time_diff)```defined above.Note: A very Pythonic way to achieve this involes zusing `zip()` and parameter unpacking using `*`. This a bit beyond scope of this excercise, a non-Pythonic way is perfectly fine for now.
###Code
calculate_growth_doubling_time(infected_cases[0], infected_cases[1], time_diff=1)
[x + 10 for x, y in [(1,2), (3,4), (5,6)] ]
list(zip(*[(1,2), (3,4), (5,6)]))
[(1,2), (3,4), (5,6)]
growth_rates, doubling_times = zip(*[calculate_growth_doubling_time(value_t1, value_t2, 1)
for value_t1, value_t2 in zip(infected_cases[:-1], infected_cases[1:])])
growth_rates, doubling_times = zip(*[calculate_growth_doubling_time(value_t1, value_t2, 1)
for value_t1, value_t2 in zip(infected_cases[:-1], infected_cases[1:])])
import matplotlib.pyplot as plt
import math
plt.plot(growth_rates, label = 'Daily growth rates')
plt.xlabel('Time')
plt.ylabel('Ratio')
plt.title('Daily growth rates')
plt.legend()
plt.show()
import matplotlib.pyplot as plt
import math
plt.plot(doubling_times, label = 'Doubling time')
plt.xlabel('Time')
plt.ylabel('Days')
plt.title('Doubling time')
plt.legend()
plt.show()
###Output
_____no_output_____
###Markdown
Exercise 6Write a function which calculates the growth rate and doubling time for all time steps, i.e. similar to the code in Exercise 5. The function is defined as```growth_rate_doubling_time(l, interval)```Interval defines how many days are between the start and the end value (i.e. value_t1 and value_t2 in calculate_growth_doubling_time). Test the function by applying it to the infection data with an interval time of 7 and plotting the growth rate and the doubling time.
###Code
def growth_rate_doubling_time(l, interval):
return zip(*[calculate_growth_doubling_time(value_t1, value_t2, interval)
for value_t1, value_t2 in zip(l[:-1], l[1:])])
# first test if the function does the same as before...
growth_rates_, doubling_times_ = growth_rate_doubling_time(infected_cases, 1)
growth_rates == growth_rates_, doubling_times == doubling_times_
# first test if the function does the same as before...
growth_rates, doubling_times = growth_rate_doubling_time(infected_cases, 7)
import matplotlib.pyplot as plt
import math
plt.plot(growth_rates, label = 'Weekly growth rates')
plt.xlabel('Time')
plt.ylabel('Ratio')
plt.title('Weekly growth rates')
plt.legend()
plt.show()
import matplotlib.pyplot as plt
import math
plt.plot(doubling_times, label = 'Doubling time')
plt.xlabel('Time')
plt.ylabel('Days')
plt.title('Doubling time')
plt.legend()
plt.show()
###Output
_____no_output_____ |
liabilities_funding_ratio.ipynb | ###Markdown
CIR Model to simulate changes in Interest Rates and Liability Hedging$$ dr_{t}=a(b-r_{t})\,dt+\sigma {\sqrt {r_{t}}}\,dW_{t} $$
###Code
def inst_to_ann(r):
"""
Converts short rate to an annualized rate
:param r:
:return:
"""
return np.expm1(r)
def ann_to_inst(r):
"""
Convert annualized to a short rate
:param r:
:return:
"""
return np.log1p(r)
%load_ext autoreload
%autoreload 2
%matplotlib inline
import pandas as pd
import numpy as np
import edhec_risk_kit as erk
import math
import matplotlib.pyplot as plt
import ipywidgets as widgets
from IPython.display import display
def cir(n_years = 10, n_scenarios=1, a=0.05, b=0.03, sigma=0.05, steps_per_year=12, r_0=None):
"""
Generate random interest rate evolution over time using the CIR model
b and r_0 are assumed to be the annualized rates, not the short rate
and the returned values are the annualized rates as well
"""
if r_0 is None: r_0 = b
r_0 = ann_to_inst(r_0)
dt = 1/steps_per_year
num_steps = int(n_years*steps_per_year) + 1 # because n_years might be a float
shock = np.random.normal(0, scale=np.sqrt(dt), size=(num_steps, n_scenarios))
rates = np.empty_like(shock)
rates[0] = r_0
## For Price Generation
h = math.sqrt(a**2 + 2*sigma**2)
prices = np.empty_like(shock)
####
def price(ttm, r):
_A = ((2*h*math.exp((h+a)*ttm/2))/(2*h+(h+a)*(math.exp(h*ttm)-1)))**(2*a*b/sigma**2)
_B = (2*(math.exp(h*ttm)-1))/(2*h + (h+a)*(math.exp(h*ttm)-1))
_P = _A*np.exp(-_B*r)
return _P
prices[0] = price(n_years, r_0)
####
for step in range(1, num_steps):
r_t = rates[step-1]
d_r_t = a*(b-r_t)*dt + sigma*np.sqrt(r_t)*shock[step]
rates[step] = abs(r_t + d_r_t)
# generate prices at time t as well ...
prices[step] = price(n_years-step*dt, rates[step])
rates = pd.DataFrame(data=inst_to_ann(rates), index=range(num_steps))
### for prices
prices = pd.DataFrame(data=prices, index=range(num_steps))
###
return rates, prices
def show_cir_prices(r_0=0.03, a=0.5, b=0.03, sigma=0.05, n_scenarios=5):
cir(r_0=r_0, a=a, b=b, sigma=sigma, n_scenarios=n_scenarios)[1].plot(legend=False, figsize=(12,5))
controls = widgets.interactive(show_cir_prices,
r_0 = (0, .15, .01),
a = (0, 1, .1),
b = (0, .15, .01),
sigma= (0, .1, .01),
n_scenarios = (1, 100))
display(controls)
a_0 = 0.75
rates, bond_prices = cir(r_0=0.03, b=0.03, n_scenarios=10, n_years=10)
liabilities = bond_prices
zc_0 = erk.pv(pd.Series(data=[1], index=[10]), r=0.03)
n_bonds = a_0/zc_0
av_zc_bonds = n_bonds*bond_prices
av_cash = a_0*(rates/12+1).cumprod()
av_cash.plot(legend=False, figsize=(12, 6))
av_zc_bonds.plot(legend=False, figsize=(12, 6))
(av_cash/av_zc_bonds).pct_change().plot(title='Returns of Funding Ratio with Cash (10 Scenarios)', legend=False,
figsize=(12, 6))
(av_zc_bonds/av_zc_bonds).pct_change().plot(title='Returns of Funding Ratio with Cash (10 Scenarios)', legend=False,
figsize=(12, 6))
a_0 = 0.6
rates, bond_prices = cir(n_scenarios=10000, r_0=0.03, b=0.03)
liabilities = bond_prices
zc_0 = erk.pv(pd.Series(data=[1], index=[10]), 0.03)
n_bonds = a_0/zc_0
av_zc_bonds = n_bonds*bond_prices
av_cash = a_0*(rates/12+1).cumprod()
tfr_cash = av_cash.iloc[-1]/liabilities.iloc[-1]
tfr_zc_bond = av_zc_bonds.iloc[-1]/liabilities.iloc[-1]
ax = tfr_cash.plot.hist(label='Cash', figsize=(15, 6), bins = 100, legend=True)
tfr_zc_bond.plot.hist(ax=ax, label="ZC Bonds", bins=100, legend=True, secondary_y=True)
###Output
_____no_output_____ |
analysis/figures/y4analysis.ipynb | ###Markdown
Octanol Learning
###Code
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/OctLearningTest-03_23_2022-12_05/YArenaInfo.mat"
df1 = loadmat(loadmat_file)['YArenaInfo']
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/OctLearningTest-03_23_2022-14_12/YArenaInfo.mat"
df2 = loadmat(loadmat_file)['YArenaInfo']
histories1 = df1['FlyChoiceMatrix'][0][0].T
schedules1 = np.concatenate([[df1['RewardStateTallyOdor1'][0][0]], [df1['RewardStateTallyOdor2'][0][0]]], axis=0).transpose((2, 1, 0))
histories2 = df2['FlyChoiceMatrix'][0][0].T[[0,2,3]]
schedules2 = np.concatenate([[df2['RewardStateTallyOdor1'][0][0][:,[0,2,3]]], [df2['RewardStateTallyOdor2'][0][0][:,[0,2,3]]]], axis=0).transpose((2, 1, 0))
histories = np.concatenate([histories1, histories2], axis=0)
schedules = np.concatenate([schedules1, schedules2], axis=0)
import seaborn as sns
sns.set(style="ticks")
sns.set(font_scale=1.2)
plt.figure(figsize=(7,7))
for n in range(7):
plt.plot(np.cumsum(histories[n]==0),np.cumsum(histories[n]==1),'-',color=plt.cm.viridis(n/6),alpha=1,linewidth=1)
plt.plot(np.cumsum(histories==0,axis=1).mean(axis=0),np.cumsum(histories==1,axis=1).mean(axis=0),linewidth=3,color='gray',label=f"Average")
plt.plot([0,len(histories[0])//2],[0,len(histories[0])//2],linewidth=2,color='black',linestyle='--')
plt.text(10,40,f"(n = {len(histories)} flies)",fontsize=14)
plt.xlabel('Cumulative number of OCT choices')
plt.ylabel('Cumulative number of MCH choices')
plt.box(False)
plt.gca().set_aspect('equal')
plt.tight_layout()
plt.savefig('OctLearningTest.png',dpi=300,transparent=True)
plt.show()
i = schedules[0]
plt.figure(figsize=(8,2))
plt.plot(np.arange(i.shape[0])[i[:,0]==1],np.zeros(np.sum(i[:,0]==1)),'o',color=plt.cm.viridis(0.6),linewidth=2)
plt.plot(np.arange(i.shape[0])[i[:,1]==1],np.ones(np.sum(i[:,1]==1)),'o',color=plt.cm.viridis(0.6),linewidth=2)
plt.plot(histories.mean(axis=0),'-',color=plt.cm.viridis(0.8),linewidth=2)
plt.yticks([0,1],["OCT","MCH"])
plt.xlim([0,i.shape[0]])
plt.axhline(0.5,linewidth=2,color='black',linestyle='--')
plt.box(False)
plt.xlabel('Trial')
plt.ylabel('Odor Choice')
plt.tight_layout()
plt.savefig('OctLearningTest-schedules.png',dpi=300,transparent=True)
plt.show()
###Output
_____no_output_____
###Markdown
Methycyclohexanol Learning
###Code
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/MchLearningTest-03_23_2022-15_50/YArenaInfo.mat"
df1 = loadmat(loadmat_file)['YArenaInfo']
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/MchLearningTest-03_24_2022-10_35/YArenaInfo.mat"
df2 = loadmat(loadmat_file)['YArenaInfo']
histories1 = df1['FlyChoiceMatrix'][0][0].T[[0,2]]
schedules1 = np.concatenate([[df1['RewardStateTallyOdor1'][0][0][:,[0,2]]], [df1['RewardStateTallyOdor2'][0][0][:,[0,2]]]], axis=0).transpose((2, 1, 0))
histories2 = df2['FlyChoiceMatrix'][0][0].T[[0,2,3]]
schedules2 = np.concatenate([[df2['RewardStateTallyOdor1'][0][0][:,[0,2,3]]], [df2['RewardStateTallyOdor2'][0][0][:,[0,2,3]]]], axis=0).transpose((2, 1, 0))
histories = np.concatenate([histories1, histories2], axis=0)
schedules = np.concatenate([schedules1, schedules2], axis=0)
import seaborn as sns
sns.set(style="ticks")
sns.set(font_scale=1.2)
plt.figure(figsize=(7,7))
for n in range(5):
plt.plot(np.cumsum(histories[n]==0),np.cumsum(histories[n]==1),'-',color=plt.cm.viridis(n/4),alpha=1,linewidth=1)
plt.plot(np.cumsum(histories==0,axis=1).mean(axis=0),np.cumsum(histories==1,axis=1).mean(axis=0),linewidth=3,color='gray',label=f"Average")
plt.plot([0,100//2],[0,100//2],linewidth=2,color='black',linestyle='--')
plt.text(40,10,f"(n = {len(histories)} flies)",fontsize=14)
plt.xlabel('Cumulative number of OCT choices')
plt.ylabel('Cumulative number of MCH choices')
plt.box(False)
plt.gca().set_aspect('equal')
plt.tight_layout()
plt.savefig('MchLearningTest.png',dpi=300,transparent=True)
plt.show()
i = schedules[0]
plt.figure(figsize=(8,2))
plt.plot(np.arange(i.shape[0])[i[:,0]==1],np.zeros(np.sum(i[:,0]==1)),'o',color=plt.cm.viridis(0.6),linewidth=2)
plt.plot(np.arange(i.shape[0])[i[:,1]==1],np.ones(np.sum(i[:,1]==1)),'o',color=plt.cm.viridis(0.6),linewidth=2)
plt.plot(histories.mean(axis=0),'-',color=plt.cm.viridis(0.8),linewidth=2)
plt.yticks([0,1],["OCT","MCH"])
plt.xlim([0,i.shape[0]])
plt.axhline(0.5,linewidth=2,color='black',linestyle='--')
plt.box(False)
plt.xlabel('Trial')
plt.ylabel('Odor Choice')
plt.tight_layout()
plt.savefig('MchLearningTest-schedules.png',dpi=300,transparent=True)
plt.show()
###Output
_____no_output_____
###Markdown
ACV Preference Experiments
###Code
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/ACVPreferenceTest-03_26_2022-10_41/YArenaInfo.mat"
df1 = loadmat(loadmat_file)['YArenaInfo']
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/ACVPreferenceTest-03_26_2022-13_00/YArenaInfo.mat"
df2 = loadmat(loadmat_file)['YArenaInfo']
histories1 = df1['FlyChoiceMatrix'][0][0].T#[[0,2]]
histories2 = df2['FlyChoiceMatrix'][0][0].T#[[0,2,3]]
histories = np.concatenate([histories1, histories2], axis=0)
import seaborn as sns
sns.set(style="ticks")
sns.set(font_scale=1.2)
plt.figure(figsize=(7,7))
for n in range(8):
plt.plot(np.cumsum(histories[n]==0),np.cumsum(histories[n]==1),'-',color=plt.cm.viridis(n/7),alpha=1,linewidth=1)
plt.plot(np.cumsum(histories==0,axis=1).mean(axis=0),np.cumsum(histories==1,axis=1).mean(axis=0),linewidth=3,color='gray',label=f"Average")
plt.plot([0,len(histories[0])//2],[0,len(histories[0])//2],linewidth=2,color='black',linestyle='--')
plt.text(40,10,f"(n = {len(histories)} flies)",fontsize=14)
plt.xlabel('Cumulative number of AIR choices')
plt.ylabel('Cumulative number of ACV choices')
plt.box(False)
plt.gca().set_aspect('equal')
plt.tight_layout()
plt.savefig('ACVPreferenceTest.png',dpi=300,transparent=True)
plt.show()
plt.figure(figsize=(8,2))
plt.plot(histories.mean(axis=0),'-',color=plt.cm.viridis(0.8),linewidth=2)
plt.yticks([0,1],["OCT","MCH"])
plt.xlim([0,i.shape[0]])
plt.axhline(0.5,linewidth=2,color='black',linestyle='--')
plt.box(False)
plt.xlabel('Trial')
plt.ylabel('Odor Choice')
plt.tight_layout()
plt.savefig('ACVPreferenceTest-schedules.png',dpi=300,transparent=True)
plt.show()
###Output
_____no_output_____
###Markdown
Reversal Experiment
###Code
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/MchLearningTest-03_24_2022-10_35/YArenaInfo.mat"
df1 = loadmat(loadmat_file)['YArenaInfo']
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/MchUnlearningOctLearningTest-03_24_2022-11_32/YArenaInfo.mat"
df2 = loadmat(loadmat_file)['YArenaInfo']
loadmat_file = "//dm11/turnerlab/Rishika/4Y-Maze/RunData/OctLearningAfterMCHUnlearningTest-03_24_2022-11_58/YArenaInfo.mat"
df3 = loadmat(loadmat_file)['YArenaInfo']
histories1 = df1['FlyChoiceMatrix'][0][0].T[[0,2,3]]
schedules1 = np.concatenate([[df1['RewardStateTallyOdor1'][0][0][:,[0,2,3]]], [df1['RewardStateTallyOdor2'][0][0][:,[0,2,3]]]], axis=0).transpose((2, 1, 0))
histories2 = df2['FlyChoiceMatrix'][0][0].T[[0,2,1]]
schedules2 = np.concatenate([[df2['RewardStateTallyOdor1'][0][0][:,[0,2,1]]], [df2['RewardStateTallyOdor2'][0][0][:,[0,2,1]]]], axis=0).transpose((2, 1, 0))
histories3 = df3['FlyChoiceMatrix'][0][0].T[[0,2,1]]
schedules3 = np.concatenate([[df3['RewardStateTallyOdor1'][0][0][:,[0,2,1]]], [df3['RewardStateTallyOdor2'][0][0][:,[0,2,1]]]], axis=0).transpose((2, 1, 0))
histories = np.concatenate([histories1, histories2, histories3], axis=1)
schedules = np.concatenate([schedules1, schedules2, schedules3], axis=1)
import seaborn as sns
sns.set(style="ticks")
sns.set(font_scale=1.2)
plt.figure(figsize=(7,7))
for n in range(3):
plt.plot(np.cumsum(histories[n]==0),np.cumsum(histories[n]==1),'-',color=plt.cm.viridis(n/2),alpha=1,linewidth=1)
plt.plot(np.cumsum(histories==0,axis=1).mean(axis=0),np.cumsum(histories==1,axis=1).mean(axis=0),linewidth=3,color='gray',label=f"Average")
plt.plot([0,len(histories[0])//2],[0,len(histories[0])//2],linewidth=2,color='black',linestyle='--')
plt.text(40,10,f"(n = {len(histories)} flies)",fontsize=14)
plt.xlabel('Cumulative number of OCT choices')
plt.ylabel('Cumulative number of MCH choices')
plt.box(False)
plt.gca().set_aspect('equal')
plt.tight_layout()
plt.savefig('ReversalLearningTest.png',dpi=300,transparent=True)
plt.show()
i = schedules[0]
plt.figure(figsize=(8,2))
plt.plot(np.arange(i.shape[0])[i[:,0]==1],np.zeros(np.sum(i[:,0]==1)),'o',color=plt.cm.viridis(0.6),linewidth=2)
plt.plot(np.arange(i.shape[0])[i[:,1]==1],np.ones(np.sum(i[:,1]==1)),'o',color=plt.cm.viridis(0.6),linewidth=2)
plt.plot(histories.mean(axis=0),'-',color=plt.cm.viridis(0.8),linewidth=2)
plt.yticks([0,1],["OCT","MCH"])
plt.xlim([0,i.shape[0]])
plt.axhline(0.5,linewidth=2,color='black',linestyle='--')
plt.box(False)
plt.xlabel('Trial')
plt.ylabel('Odor Choice')
plt.tight_layout()
plt.savefig('ReversalLearningTest-schedules.png',dpi=300,transparent=True)
plt.show()
###Output
_____no_output_____ |
SentinelHub/.ipynb_checkpoints/Untitled-checkpoint.ipynb | ###Markdown
script from https://mygeoblog.com/2017/10/06/from-gee-to-numpy-to-geotiff/
###Code
# get data into different arrays
#data = np.array((ee.Array(latlon.get("nd")).getInfo()))
B1 = np.array((ee.Array(latlon.get('B1')).getInfo()))
B2 = np.array((ee.Array(latlon.get('B2')).getInfo()))
B4 = np.array((ee.Array(latlon.get('B4')).getInfo()))
B5 = np.array((ee.Array(latlon.get('B5')).getInfo()))
B8 = np.array((ee.Array(latlon.get('B8')).getInfo()))
B8A = np.array((ee.Array(latlon.get('B8A')).getInfo()))
B9 = np.array((ee.Array(latlon.get('B9')).getInfo()))
B10 = np.array((ee.Array(latlon.get('B10')).getInfo()))
B11 = np.array((ee.Array(latlon.get('B11')).getInfo()))
B12 = np.array((ee.Array(latlon.get('B12')).getInfo()))
lats = np.array((ee.Array(latlon.get("latitude")).getInfo()))
lons = np.array((ee.Array(latlon.get("longitude")).getInfo()))
# get the unique coordinates
uniqueLats = np.unique(lats)
uniqueLons = np.unique(lons)
# get number of columns and rows from coordinates
ncols = len(uniqueLons)
nrows = len(uniqueLats)
# determine pixelsizes
ys = uniqueLats[1] - uniqueLats[0]
xs = uniqueLons[1] - uniqueLons[0]
data = np.array((B1,B2,B4,B5,B8,B8A,B9,B10,B11,B12))
#for i in range(0,10):
# print( i )
# create an array with dimensions of image (3D array, as it is a multiband image we want)
# this has not worked properly - check script
arr2 = np.zeros([nrows, ncols,10], np.float32) #-9999
counter2 =0
for y in range(0,len(arr2),1):
for x in range(0,len(arr2[0]),1):
for z in range(0,10): #put 10 here, as 10 bands are input (put 13 if all 13 are input etc)
if lats[counter2] == uniqueLats[y] and lons[counter2] == uniqueLons[x] and counter2 < len(lats)-1:
counter2+=1
arr2[len(uniqueLats)-1-y,x] = data[z][counter2] # we start from lower left corner
import matplotlib.pyplot as plt
plt.imshow(arr2[:,:,1])
plt.show()
arr2[:,:,5]
cloud_detector = S2PixelCloudDetector(threshold=0.4, average_over=4, dilation_size=2)
cloud_probs = cloud_detector.get_cloud_probability_maps(arr2)
cloud_masks = cloud_detector.get_cloud_masks(np.array(wcsbands))
###Output
_____no_output_____ |
docs/examples/xopt_parallel.ipynb | ###Markdown
Xopt Parallel ExamplesXopt provides methods to parallelize optimizations using Processes, Threads, MPI, and Dask using the `concurrent.futures` interface as defined in https://www.python.org/dev/peps/pep-3148/ .
###Code
# Import the class
from xopt import Xopt
# Notebook printing output
#from xopt import output_notebook
#output_notebook()
!mkdir temp
###Output
_____no_output_____
###Markdown
The `Xopt` object can be instantiated from a JSON or YAML file, or a dict, with the proper structure.Here we will make one
###Code
# Make a proper input file.
YAML="""
xopt:
output_path: temp
algorithm:
name: cnsga
options:
max_generations: 5
population_size: 128
show_progress: True
simulation:
name: test_TNK
evaluate: xopt.tests.evaluators.TNK.evaluate_TNK
vocs:
variables:
x1: [0, 3.14159]
x2: [0, 3.14159]
objectives: {y1: MINIMIZE, y2: MINIMIZE}
constraints:
c1: [GREATER_THAN, 0]
c2: [LESS_THAN, 0.5]
linked_variables: {x9: x1}
constants: {a: dummy_constant}
"""
###Output
_____no_output_____
###Markdown
Processes
###Code
from concurrent.futures import ProcessPoolExecutor
X = Xopt(YAML)
with ProcessPoolExecutor() as executor:
X.run(executor=executor)
###Output
_____no_output_____
###Markdown
ThreadsContinue running, this time with threads.
###Code
from concurrent.futures import ThreadPoolExecutor
with ThreadPoolExecutor() as executor:
X.run(executor=executor)
###Output
_____no_output_____
###Markdown
MPI The `test.yaml` file completely defines the problem. We will also direct the logging to an `xopt.log` file. The following invocation recruits 4 MPI workers to solve this problem.We can also continue by calling `.save` with a JSON filename. This will write all of previous results into the file.
###Code
# Write YAML text to a file
# open('test.yaml', 'w').write(YAML)
X.save('test.json')
!mpirun -n 4 python -m mpi4py.futures -m xopt.mpi.run -vv --logfile xopt.log test.json
###Output
Namespace(input_file='test.json', logfile='xopt.log', verbose=2)
Parallel execution with 4 workers
Loading from JSON file: test.json
Loading config from dict.
Loading config from dict.
Loading config from dict.
Loading config from dict.
Specified both known algorithm `cnsga` and `function`. Using known algorithm function.
Starting at time 2021-10-08T13:13:09-07:00
▄████▄ ███▄ █ ██████ ▄████ ▄▄▄
▒██▀ ▀█ ██ ▀█ █ ▒██ ▒ ██▒ ▀█▒▒████▄
▒▓█ ▄ ▓██ ▀█ ██▒░ ▓██▄ ▒██░▄▄▄░▒██ ▀█▄
▒▓▓▄ ▄██▒▓██▒ ▐▌██▒ ▒ ██▒░▓█ ██▓░██▄▄▄▄██
▒ ▓███▀ ░▒██░ ▓██░▒██████▒▒░▒▓███▀▒ ▓█ ▓██▒
░ ░▒ ▒ ░░ ▒░ ▒ ▒ ▒ ▒▓▒ ▒ ░ ░▒ ▒ ▒▒ ▓▒█░
░ ▒ ░ ░░ ░ ▒░░ ░▒ ░ ░ ░ ░ ▒ ▒▒ ░
░ ░ ░ ░ ░ ░ ░ ░ ░ ░ ░ ▒
░ ░ ░ ░ ░ ░ ░
░
Continuous Non-dominated Sorting Genetic Algorithm
Version 0.4.3+219.gfff1660.dirty
Creating toolbox from vocs.
Created toolbox with 2 variables, 2 constraints, and 2 objectives.
Using selection algorithm: nsga2
Loading config from dict.
Initializing with existing population, size 128
Maximum generations: 11
____________________________________________________
128 fitness calculations for initial generation
done.
Submitting first batch of children
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/initial_pop_6.json
129it [00:00, 2822.75it/s]
Generation 6: 0%| | 0/128 [00:00<?, ?it/s]Generation 6 completed in 0.00117 minutes
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/gen_7.json
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/pop_7.json
Generation 6: 100%|██████████| 128/128 [00:00<00:00, 715.93it/s]
Generation 7: 0%| | 0/128 [00:00<?, ?it/s]Generation 7 completed in 0.00299 minutes
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/gen_8.json
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/pop_8.json
Generation 7: 100%|██████████| 128/128 [00:00<00:00, 767.73it/s]
Generation 8: 0%| | 0/128 [00:00<?, ?it/s]Generation 8 completed in 0.00279 minutes
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/gen_9.json
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/pop_9.json
Generation 8: 100%|██████████| 128/128 [00:00<00:00, 806.00it/s]
Generation 9: 0%| | 0/128 [00:00<?, ?it/s]Generation 9 completed in 0.00265 minutes
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/gen_10.json
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/pop_10.json
Generation 9: 100%|██████████| 128/128 [00:00<00:00, 816.36it/s]
Generation 10: 0%| | 0/128 [00:00<?, ?it/s]Generation 10 completed in 0.00262 minutes
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/gen_11.json
Pop written to /Users/chrisonian/Code/GitHub/xopt/examples/basic/temp/pop_11.json
Generation 10: 130it [00:00, 952.17it/s]
###Markdown
Dask
###Code
from dask.distributed import Client
with Client() as executor:
X.run(executor=executor)
###Output
_____no_output_____
###Markdown
Load output into PandasThis algorithm writes two types of files: `gen_{i}.json` with all of the new individuals evaluated in a generation, and `pop_{i}.json` with the latest best population. Xopt provides some functions to load these easily into a Pandas dataframe for further analysis.
###Code
from xopt.dataset import load_all_xopt_data
from glob import glob
%config InlineBackend.figure_format = 'retina'
# Get a list of all of the gen files
genfiles = glob('temp/gen_*json')
df = load_all_xopt_data(genfiles)
df
# Plot the feasible ones
feasible_df = df[df['feasible']]
feasible_df.plot('y1', 'y2', kind='scatter').set_aspect('equal')
# Plot the infeasible ones
infeasible_df = df[~df['feasible']]
infeasible_df.plot('y1', 'y2', kind='scatter').set_aspect('equal')
# This is the final population
df1 = load_all_xopt_data(['temp/pop_17.json'])
df1.plot('y1', 'y2', kind='scatter').set_aspect('equal')
###Output
_____no_output_____
###Markdown
matplotlib plottingYou can always use matplotlib for customizable plotting
###Code
import matplotlib.pyplot as plt
%matplotlib inline
# Extract objectives from output
k1, k2 = 'y1', 'y2'
fig, ax = plt.subplots(figsize=(6,6))
ax.scatter(infeasible_df[k1], infeasible_df[k2],color='blue', marker='.', alpha=0.5, label='infeasible')
ax.scatter(feasible_df[k1], feasible_df[k2],color='orange', marker='.', label='feasible')
ax.scatter(df1[k1], df1[k2],color='red', marker='.', label='final population')
ax.set_xlabel(k1)
ax.set_ylabel(k2)
ax.set_aspect('auto')
ax.set_title(f" {X.simulation['name']} using Xopt's CNSGA algorithm" )
plt.legend()
# Cleanup
!rm -r dask-worker-space
!rm -r temp
!rm xopt.log
!rm test.json
###Output
_____no_output_____ |
assignments/homework_5.ipynb | ###Markdown
Homework Assignment 5 - Logistic Regression, MLP, and CNN with Torch In this assignment, we will solve the face recognition problem again. This time we will be using a multilayer perceptron (MLP) and a convolutional neural network (CNN). We will do so usign the PyTorch framework. Reminders- Start by making a copy of this notebook in order to be able to save it.- Use **Ctrl+[** to expend all cells. Tip of the day - Progress BarWhen running a long calculation, we would usually want to have a progress bar to track the progress of our process. One great python package for creating such a progress bar is [**tqdm**](https://github.com/tqdm/tqdm). This package is easy to use and offers a highly customizable progress bar. For example, to add a progress bar to an existing loop, simply surrounding the iterable which the loops run over with the **tqdm_notebook** command:```pythonimport tqdmfor x in tqdm.tqdm_notebook(some_list): some_long_running_function(x)```✍️ Add a progress bar to the following loop:
###Code
import tqdm
import time
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
for i in range(10):
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
print('Step {}'.format(i))
time.sleep(1)
###Output
_____no_output_____
###Markdown
Your IDs✍️ Fill in your IDs in the cell below:
###Code
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
# Replace the IDs bellow with our own
student1_id = '012345678'
student2_id = '012345678'
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
print('Hello ' + student1_id + ' & ' + student2_id)
###Output
_____no_output_____
###Markdown
Importing PackagesImporting the NumPy, Pandas and Matplotlib packages.
###Code
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
## This line makes matplotlib plot the figures inside the notebook
%matplotlib inline
## Set some default values of the the matplotlib plots
plt.rcParams['figure.figsize'] = (8.0, 8.0) # Set default plot's sizes
plt.rcParams['axes.grid'] = True # Show grid by default in figures
###Output
_____no_output_____
###Markdown
Graphical Processing Unit (GPU)GPUs are special processing cards which were originally developed to for accelerating graphical related calculations such as 3D rendering and adujsting and playing a high-resolution video. Today these cards are also in use for a variety of tasks which are not necessarily graphic related, such as training neural networks.These GPUs are optimal for parallelizing simple operations on a large amount of data. The CPU (Central Processing Unit) is the computer's main processing unit and usually has a few fast and "strong" processing components called cores (usually up to tens of cores ). As opposed to it, a GPUs usually has many (usually thousands of) slower and "weaker" cores.When running a process which performs some mathematical operation to a large amount of data, for example calculating the $e^x$ for each element in a large matrix or multiplication between two matrices, we can speed up our process significantly by running it on a GPU.The GPU does not share the same memory space with the CPU and has its own memory. Therefore before performing any calculation using a GPU, we must first transfer our data to the GPU's memory. Colab and GPUsIn this assignment, we will train our network on a GPU. Colab offers free GPU support, but it is not enabled by default. To enable it, go to the menu bar, open **Runtime->Change runtime type** and change **hardware accelerator** to GPU. Click **save** to save your selection. You will see how we can tell our code to perform an operation on the GPU instead of on the CPU. PyTorchPyTorch is a framework (a collection of tools) which significantly simplify the process of building and training neural networks. This framework was initially developed and is currently backed by Facebook. PyTorch is only one of many great such frameworks which currently exist. For a list of some of the commonly used frameworks today, see workshop 9.Specifically, in this assignment, we will relay on PyTorch for the following features:- The package's ability to automatic calculate gradients (using back-propagation)- The package's ability to move a variable to the GPU and perform calculations on it.- The package's stochastic gradient descent optimization object.- The built-in objects/function for building and training models: - Linear layer - Convolutional layers - Relu - SoftMax - Minus-logLikelihood loss In this homework assignment, we will **not** cover all of what is needed for using PyTorch. It is aimed to show you the basics idea of what the framework has to offer. To better understand PyTorch, a good place to start are the great tutorials on the package's [website](https://pytorch.org/tutorials/index.html). The "60 Minute Blitz" along with the "Learning PyTorch with Examples" on the website provide a great starting point. TensorsThe basic PyTorch object is the tensor with has a very similar (but not exact) interface to that of the NumPy array. A few differences which are worth mentioning:- Tensors do not yet support the "@" operator of performing matrix multiplication. It is performed by using **torch.matmul(a_mat, b_mat)**.- The transpose of a matrix is given by **a_mat.t()** (instead of the **a_mat.T** method which is in use in numpy)For example:
###Code
import torch ## importing PyTorch
## Defining a tensor from lists of numbers
x1 = torch.tensor([[1.,2.], [3., 4.]])
print('x1=\n{}\n'.format(x1))
## Creating a random tensor
x2 = torch.randint(low=0, high=10, size=(2, 3)).float()
print('x2=\n{}\n'.format(x2))
## Multipliing tensors
y = torch.matmul(x1.t(), x2)
print('torch.matmul(x1.t(), x2)=\n{}'.format(y))
###Output
_____no_output_____
###Markdown
An Important Comment About Single & Double Precision & Fixed PointsBy default, numpy uses 64 bit to store floating point numbers. This representation is optimal for most CPUs. In contrast to that, PyTorch uses 32 bits, which is optimal for most GPUs. The 64-bit representation is called **double precision**, and the 32-bit is called **single precision**.Most of PyTorch's operations can only be performed only between two tensors of the same type. Therefore we will make sure that all of our tensors will be stored using single precision. You can convert a tensor to single precision representation by using the tensors **.float()** command.For some of the operations, we will also need to convert fixed point tensors (integers) to single precision. This is done in a similar way using the **.float()** command.For example:
###Code
## This is a fixed point tensor
x_int = torch.tensor([4, 2, 3])
print('x_int=\n{}'.format(x_int))
print('x_int.dtype={}\n'.format(x_int.dtype))
## Converting the tensor to single persicion
x_single = x_int.float()
print('x_single=\n{}'.format(x_single))
print('x_single.dtype={}\n'.format(x_single.dtype))
## Converting the tensor to double persicion
x_doubel = x_int.double()
print('x_doubel=\n{}'.format(x_doubel))
print('x_doubel.dtype={}\n'.format(x_doubel.dtype))
###Output
_____no_output_____
###Markdown
PyTorch and GPUsPyTorch provides a simple way to copy a tensor to the GPU's memory. The GPUs In PyTorch are referred to as [**CUDA**](https://en.wikipedia.org/wiki/CUDA) devices. A tensor can be copied to the GPU's memory by using the tensor's **.cuda** command. All the mathematical operations can then be performed on the copied tensor in the same way as if it was in the regular memory. The result of a calculation which was performed on the GPU will be stored on the GPU's memory as well.A tensor can be copied back from the GPU's memory to the regular memory using the **.cpu** command. For example:
###Code
## Moving x1 to the GPU
x1_gpu = x1.cuda()
print('x1_gpu=\n{}\n'.format(x1_gpu))
## Moving x2 to the GPU
x2_gpu = torch.randint(low=0, high=10, size=(2, 3)).float().cuda()
print('x2_gpu=\n{}\n'.format(x1_gpu))
## Performing matrix multiplication on the GPU
y = torch.matmul(x1_gpu.t(), x2_gpu)
print('torch.matmul(x1_gpu.t(), x2_gpu)=\n{}\n'.format(y))
print('y.cpu()=\n{}'.format(y.cpu()))
###Output
_____no_output_____
###Markdown
Notice the **device='cuda:0'** which is attached to the outputs of tensors which are stored on the GPU's memory. ✍️ Calculate the multiplication table (לוח הכפל ($\left[1,2,\ldots,10\right]^T\left[1,2,\ldots,10\right]$)) on the GPU, copy the result back to the CPU and print the result:- PYTorch cannot multiply fixed point tensor on the GPU. Therefore so make sure you convert the tensors to single precision tensors.
###Code
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
mult_table = ...
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
print(mult_table)
###Output
_____no_output_____
###Markdown
Calculating gradientsOne of PyTorch's main features is its ability to automatically calculate gradients by using back-propagation.To calculate the gradient of a function, we need to preforme the following steps:1. Select the variables according to which we would want to calculate the derivative.2. Clear all previous gradient calculations.3. Calculate the result of the functions for a given set of variables. (the forward path)4. Run the back-propagation function starting from the calculated result of the function.Let us start with an example, and then explain it.The following code calculates the following derivative: $\left.\frac{\partial}{\partial x}x^2+5x+4\right|_{x=3}$:
###Code
## Define the variables which we would want to calculate the derivative according to
x = torch.tensor(3).float()
x.requires_grad = True
## Calculate the function's result
y = x ** 2 + 5 * x + 4
## Run back-propagation
y.backward()
## Prin the result
x_grad = x.grad
print('The derivative is: {}'.format(x_grad))
###Output
_____no_output_____
###Markdown
In the above cell, we have performed the following steps:1. We have first defined a tensor **x**, and then marked be setting it's **.requires_grad** field to **True**. This tells PyTorch that we will later want to calculate the derivative according to it.2. We have calculated the function's result (this is the forward path).3. We have used the result of the function to initiate the back-propagation calculation by using the **.backword()** function of the result tensor.After the back-propagation step, the derivative of the function according to each one of the selected variables will be stored in the **.grad** field of each of the variables.In this case, we did not have to clear any previous calculation since we did yet run any backward calculation using these variables. ✍️ Calculate and plot the derivative of the sigmoid function $\frac{1}{1+e^{-x}}$
###Code
vals = np.arange(-10, 10, 0.1)
res = np.zeros_like(vals)
for i in range(len(vals)):
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
x = torch.tensor(vals[i]).float()
...
res[i] = x.grad
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
fig, ax = plt.subplots()
ax.plot(vals, res);
ax.set_title('$\\frac{\\partial}{\\partial x}\\frac{1}{1+e^{-x}}$', fontsize=20)
ax.set_xlabel('$x$');
###Output
_____no_output_____
###Markdown
Preparing the dataRepeat the preparation of the data in the same manner as in the last assignment, but this time **do not add the additional constant 1** to the features vector. ✍️ Complete the code below to load the data, split it, and extract the PCA features.
###Code
from sklearn.datasets import fetch_lfw_people
def load_lfw_dataset():
"""
Loading the Labeled faces in the Wild dataset.
Load only face of persons which appear at least 50 times in the dataset.
Using:
- N: The number of samples in the dataset.
- H: the images' height
- W: the images' width
- K: The number of classes.
Returns
-------
x: ndarray
The N x H x W array of images.
y: ndarray
The 1D array of length N of labels.
n_classes: int
The number of different classes, K.
label_to_name_mapping: list
A list of K strings containing the name related to each label.
image_shape: list
The image's shape as the list: [H, W]
"""
dataset = fetch_lfw_people(min_faces_per_person=50)
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
...
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
return x, y, n_classes, label_to_name_mapping, image_shape
x, y, n_classes, label_to_name_mapping, image_shape = load_lfw_dataset()
print('Number of images in the dataset: {}'.format(len(x)))
print('Number of different persons in the dataset: {}'.format(n_classes))
print('Each images size is: {}'.format(image_shape))
_, images_per_class = np.unique(y, return_counts=True)
fig, ax = plt.subplots()
ax.bar(label_to_name_mapping, images_per_class)
ax.set_xticklabels(label_to_name_mapping, rotation=-90);
ax.set_title('Images per person')
ax.set_ylabel('Number of images')
fig, ax_array = plt.subplots(4, 5)
for i, ax in enumerate(ax_array.flat):
ax.imshow(x[i], cmap='gray')
ax.set_ylabel(label_to_name_mapping[y[i]])
ax.set_yticks([])
ax.set_xticks([])
def split_dataset(x, y, train_fraction=0.6, validation_fraction=0.2):
"""
Split the data
Parameters
----------
x: ndarray
The N x H x W array of images.
y: ndarray
The 1D array of length N of labels.
train_fraction: float
The fraction of the dataset to use as the train set.
validation_fraction: float
The fraction of the dataset to use as the validation set.
Returns
-------
n_samples_train: int
The number of train samples.
x_train: ndarray
The n_samples_train x H x W array of train images.
y_train: ndarray
The 1D array of length n_samples_train of train labels.
n_samples_val: int
The number of validation samples.
x_val: ndarray
The n_samples_val x H x W array of validation images.
y_val: ndarray
The 1D array of length n_samples_val of validation labels.
n_samples_test: int
The number of test samples.
x_test: ndarray
The n_samples_test x H x W array of test images.
y_test: ndarray
The 1D array of length n_samples_test of test labels.
"""
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
## Create a random generator using a fixed seed
rand_gen = np.random.RandomState(0)
...
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
return n_samples_train, x_train, y_train, n_samples_val, x_val, y_val, n_samples_test, x_test, y_test
n_samples_train, x_train, y_train, n_samples_val, x_val, y_val, n_samples_test, x_test, y_test = split_dataset(x, y)
print('Number of training samples: {}'.format(n_samples_train))
print('Number of validation samples: {}'.format(n_samples_val))
print('Number of test samples: {}'.format(n_samples_test))
from sklearn.decomposition import PCA
def generate_pca_object(x_train, n_pca_components=300):
"""
Generate a training sklearn.decomposition.PCA object.
Using:
- N: The number of samples in x_train.
Parameters
----------
x_train: ndarray
The N x H x W array of train images.
n_pca_components: int
The number of PCA components to use.
Returns
-------
pca: sklearn.decomposition.PCA
The trained sklearn.decomposition.PCA object which can perform the PCA decomposition and reconstruction
"""
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
...
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
return pca
def extract_features(x, pca):
"""
Extract features from the images, which include the PCA elements and the constant 1.
Using:
- N: The number of samples in x.
- D_PCA: The number of components.
Parameters
----------
x: ndarray
The N x H x W array of images (x can be either the train, validation, or test dataset).
pca: sklearn.decomposition.PCA
The trained sklearn.decomposition.PCA object which can perform the PCA decomposition and reconstruction
Returns
-------
features: ndarray
The N x (D_PCA + 1) array of features.
"""
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
...
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
return features
pca = generate_pca_object(x_train)
features_train = extract_features(x_train, pca)
features_val = extract_features(x_val, pca)
features_test = extract_features(x_test, pca)
## Ploting the reconstruction of the first 10 test images
reconstructed_images_flat = pca.inverse_transform(features_test)
reconstructed_images = reconstructed_images_flat.reshape(-1, *image_shape)
fig, ax_array = plt.subplots(2, 10, figsize=(15, 4))
for i in range(10):
ax_array[0][i].imshow(x_test[i], cmap='gray')
ax_array[0][i].set_yticks([])
ax_array[0][i].set_xticks([])
ax_array[1][i].imshow(reconstructed_images[i], cmap='gray')
ax_array[1][i].set_yticks([])
ax_array[1][i].set_xticks([])
ax_array[0][0].set_ylabel('Original')
ax_array[1][0].set_ylabel('Reconstructed')
fig.suptitle('Reconstructed image');
###Output
_____no_output_____
###Markdown
We will now convert all the data to tensors, transfer it to the GPU and to single precision:
###Code
x_train_gpu = torch.tensor(x_train).float().cuda()
features_train_gpu = torch.tensor(features_train).float().cuda()
y_train_gpu = torch.tensor(y_train).cuda()
x_val_gpu = torch.tensor(x_val).float().cuda()
features_val_gpu = torch.tensor(features_val).float().cuda()
y_val_gpu = torch.tensor(y_val).cuda()
x_test_gpu = torch.tensor(x_test).float().cuda()
features_test_gpu = torch.tensor(features_test).float().cuda()
y_test_gpu = torch.tensor(y_test).cuda()
###Output
_____no_output_____
###Markdown
Logistic RegressionWe will start by returning to the logistic regression model from last assignment.**Reminder**: We are modeling the conditional distribution of the labels as:$$p\left(y|\boldsymbol{x};\Theta\right)=\sigma\left(\Theta\boldsymbol{x},y\right)$$And our objective is to minimize the log likelihood of this probability:$$\boldsymbol{\theta}^* = \underset{\boldsymbol{\theta}}{\arg\min}-\frac{1}{N}\sum_i\log\left(\sigma\left(\Theta\boldsymbol{x}_i,y_i\right)\right)$$Where $\sigma$ is the softmax function:$$\sigma\left(\boldsymbol{q},k\right)=\frac{e^{q_k}}{\sum_{k'} e^{q_{k'}}}$$✍️ Complete the code below to define the same objective function we defined last time:$$g\left(\Theta;X,\boldsymbol{y}\right)=-\frac{1}{N}\sum_i\log\left(\sigma\left(\Theta\boldsymbol{x}_i,y_i\right)\right)$$This time, implement the function by using the two following function from torch:- [**torch.nn.functional.log_softmax**](https://pytorch.org/docs/stable/nn.htmltorch.nn.functional.log_softmax): which calculates the log of the softmax function. This function is similar to the softmax function which you have implemented in the last assignment with the addition of taking the log of the result.- [**torch.nn.functional.nll_loss**](https://pytorch.org/docs/stable/nn.htmltorch.nn.functional.nll_loss): Which calculates the negative log-likelihood of a matrix of log-probabilities, $P$, and a vector of labels, $\boldsymbol{y}$: $-\frac{1}{N}\sum_iP_{i,y_i}$
###Code
## Define the objective function
def g(theta, x, y):
"""
The objective function.
Using:
- N: The number of samples.
- D: The number of features (n_pca_componens + 1).
- K: The number of classes.
Parameters
----------
theta: ndarray
The K x D parameters matrix.
x: ndarray
The N x D features matrix.
y: ndarray
The 1D array of length N of labels.
Returns
-------
res: float
The objective function evaluated at the given theta.
"""
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
...
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
return res
## Testing the function
test_theta = torch.tensor([[1, -3, 2], [-2, 1, -1]]).float().cuda()
test_x = torch.tensor([[0.1, 0.7, -0.2], [0.5, -0.2, 0.5]]).float().cuda()
test_y = torch.tensor([1, 0]).cuda()
print(float(g(test_theta, test_x, test_y)))
###Output
_____no_output_____
###Markdown
Make sure the result you get is: $0.033094\ldots$ Doing it the PyTorch way - Using Torch.nn.Sequential and a Loss FunctionAn alternative way to define the model function, is by using the **[torch.nn.Sequential](https://pytorch.org/docs/stable/nn.htmltorch.nn.Sequential)** operator. The Senquential operator allows us to define a model by stacking together a chain of operators. This method is usefull when stacking layers of a neural-network model, as we will see later.In addition to the model, we need to also define our loss function between the output of the model and the labels. PyTorch offers a large variety of [such loss functions](https://pytorch.org/docs/stable/nn.htmlloss-functions).In the case of logistic regression, the model is simply a linear (fully connected) layer followed by a softmax operation. The loss function in our case is the minus log-likelihood function. Therefore, we can calculate the model in the following way:
###Code
model = torch.nn.Sequential(
torch.nn.Linear(in_features=3, out_features=2, bias=False),
torch.nn.LogSoftmax(dim=1),
).cuda()
loss_func = torch.nn.NLLLoss()
## Testing the function
model[0].weight[:] = torch.tensor([[1, -3, 2], [-2, 1, -1]]).float().cuda()
test_x = torch.tensor([[0.1, 0.7, -0.2], [0.5, -0.2, 0.5]]).float().cuda()
test_y = torch.tensor([1, 0]).cuda()
log_prob = model(test_x)
loss = loss_func(log_prob, test_y)
print(float(loss))
###Output
_____no_output_____
###Markdown
In the above code, we used have used the torch.nn.Sequential to stack together a [**torch.nn.Linear**](https://pytorch.org/docs/stable/nn.htmltorch.nn.Linear) layer and a [**torch.nn.LogSoftmax**](https://pytorch.org/docs/stable/nn.htmltorch.nn.LogSoftmax) layer. The linear layer is defined by the number of **input_features**, the number of **output_features** and a flag for optionally added a bias term.The log-softmax layer is defined by the dimension in which the softmax is calculated.Notice that we have used here [**torch.nn.LogSoftmax**](https://pytorch.org/docs/stable/nn.htmltorch.nn.LogSoftmax) to define an operator (rather then [**torch.nn.functional.log_softmax**](https://pytorch.org/docs/stable/nn.htmltorch.nn.functional.log_softmax) which is a function)In addition, notice that we have used the **.cuda** function to copy all the model parameters to the GPU.The parameters of each operator are automatically defined for each operator (which in this case, is the matrix $\Theta$ of the linear layer). They are stored as part of the model inside each of the operators. The Gradient Decent AlgorithmWe will use the following function for running the gradient descent algorithm with an L2 regularization.A few points which are worth mentioning regarding the code:1. It uses the built-in optimization object [**torch.optim.SGD**](https://pytorch.org/docs/stable/optim.htmltorch.optim.SGD) for performing the gradient step.2. We are using the weight decay option of torch.optim.SGD which is equivalet to adding an L2 regularization to the objective.
###Code
def train_model(model, alpha, n_iters, x_train, y_train, x_val, y_val, llambda):
## Initialize lists to store intermidiate results for plotting
objective_list_train = []
objective_list_val = []
## Defining the loss function
loss_func = torch.nn.NLLLoss()
## Defein Optimizer
optimizer = torch.optim.SGD(params=model.parameters(), lr=alpha, weight_decay=2 * llambda)
## Perforing the update steps
for i_iter in tqdm.tqdm_notebook(range(n_iters)):
## reseting all previous gradients
optimizer.zero_grad()
## Forward path
log_prob = model(x_train)
log_prob = log_prob.view(log_prob.shape[0], log_prob.shape[1]) ## This is for later support for the CNNs
loss = loss_func(log_prob, y_train)
## Backward path
loss.backward()
## Optimization step
optimizer.step()
## Store intermidiate results
objective_list_train.append(float(loss))
with torch.no_grad(): ## This line is important and it tells PyTorch nit to calculate gradiants in this section
log_prob = model(x_val)
log_prob = log_prob.view(log_prob.shape[0], log_prob.shape[1]) ## This is for later support for the CNNs
loss = loss_func(log_prob, y_val)
objective_list_val.append(float(loss))
objectives_array_train = np.array(objective_list_train)
objectives_array_val = np.array(objective_list_val)
return objectives_array_train, objectives_array_val
###Output
_____no_output_____
###Markdown
Training✍️ Complete the code below to define the logistig regression model and used the above function to train it:
###Code
n_features = features_train_gpu.shape[1]
## The logistic regression model
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
model = torch.nn.Sequential(
...
).cuda()
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
## To save you time, some optimal hyper-parameters were pre-selected.
alpha = 1e-5 ## Learning rate parameter
llambda = 100 ## L2 regularization parameter
n_iters = 2000
objectives_array_train, objectives_array_val = train_model(model, alpha, n_iters,
features_train_gpu, y_train_gpu,
features_val_gpu, y_val_gpu,
llambda)
## Plot the objective
## ==================
fig, ax = plt.subplots()
ax.plot(objectives_array_train, label='Train')
ax.plot(objectives_array_val, label='Validation')
ax.set_title('Otimization objective')
ax.set_xlabel('Step')
ax.set_xlabel('Objective')
ax.set_ylim(0, 5)
ax.legend();
with torch.no_grad():
y_hat_test = model(features_test_gpu).argmax(dim=1)
empirical_risk_test = (y_hat_test != y_test_gpu).float().mean()
print('The empirical risk (amount of misclassifications) on the test set is: {}'.format(empirical_risk_test))
## Plot estimation
## ===============
fig, ax_array = plt.subplots(4, 5)
for i, ax in enumerate(ax_array.flat):
ax.imshow(x_test[i], cmap='gray')
ax.set_yticks([])
ax.set_xticks([])
ax.set_ylabel(label_to_name_mapping[y_hat_test[i]].split()[-1],
color='black' if y_hat_test[i] == y_test[i] else 'red')
fig.suptitle('Predicted Names; Incorrect Labels in Red', size=14);
###Output
_____no_output_____
###Markdown
MLPNow that we have a training function and we know how to define models using PyTorch, we can start playing around with some neural-networks architectures. Specifically, we will run one MLP network and one CNN network.✍️ Complete the code below to define an MLP with 1 hidden layer of 1024 neurons and a ReLU activation function.I.e., build a network which with of the following layers:1. A fully connected (linear) layer with an input of the n_features and output of 1024.2. A ReLU layer3. A fully connected (linear) layer with an input of 1024 and output of n_classes.4. A log-softmax function.- Use [**torch.nn.ReLU**(https://pytorch.org/docs/stable/nn.htmltorch.nn.ReLU)] to define the ReLU layer.
###Code
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
model = torch.nn.Sequential(
...
).cuda()
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
## We will use the following
alpha = 1e-3
llambda = 0.1
n_iters = 20000
objectives_array_train, objectives_array_val = train_model(model, alpha, n_iters,
features_train_gpu, y_train_gpu,
features_val_gpu, y_val_gpu,
llambda)
## Plot the objective
## ==================
fig, ax = plt.subplots()
ax.plot(objectives_array_train, label='Train')
ax.plot(objectives_array_val, label='Validation')
ax.set_title('Otimization objective')
ax.set_xlabel('Step')
ax.set_xlabel('Objective')
ax.set_ylim(0, 5)
ax.legend();
with torch.no_grad():
y_hat_test = model(features_test_gpu).argmax(dim=1)
empirical_risk_test = (y_hat_test != y_test_gpu).float().mean()
print('The empirical risk (amount of misclassifications) on the test set is: {}'.format(empirical_risk_test))
###Output
_____no_output_____
###Markdown
Make sure you get a test risk of about 16%. CNNAs opposed to using the PCA features as an input to our mode,l we can use the raw images directly, but for that, we will need a different architecture.✍️ Complete the code below to define a CNN which is composed of the following layers:1. A convolutional layer with a 4x4 kernel, 64 output channels, a stride of 2, and a padding of 2 on the vertical direction and 4 on the horizontal direction.2. A ReLU layer.3. A convolutional layer with a 4x4 kernel, 128 output channels, a stride of 2, and a padding of 1 in each direction.4. A ReLU layer.5. A convolutional layer with a 4x4 kernel, 256 output channels, a stride of 2, and a padding of 1 in each direction.6. A ReLU layer.7. A convolutional layer with a 4x4 kernel, 512 output channels, a stride of 2, and a padding of 1 in each direction.8. A ReLU layer.9. A convolutional layer (which is also a fully connected layer) with a 4x3 kernel and n_classes output channels (with no padding)10. A log-softmax layer.- Use [**torch.nn.MaxPool2d**(https://pytorch.org/docs/stable/nn.htmltorch.nn.MaxPool2d)] to define the max pooling layers. Oprn the documentation to check the function parameters.- Implement the fully connected layer using a convolutional layer with a kernel of size 1,- The learning rate, which was pre-selected for training the network, is a bit too high for the task and was chosen to produce reasonable results in reasonable time. Training the network with the given number of iteration should take about 5 minutes. - This architecture is very far from being optimal and was built to be simple, and with a reasonable training time.
###Code
## %%%%%%%%%%%%%%% Your code here - Begin %%%%%%%%%%%%%%%
model = torch.nn.Sequential(
torch.nn.Conv2d(in_channels=1, out_channels=64, kernel_size=4, stride=2, padding=(2, 4) ),
...
).cuda()
## %%%%%%%%%%%%%%% Your code here - End %%%%%%%%%%%%%%%%%
alpha = 1e-3
llambda = 0.1
n_iters = 1000
objectives_array_train, objectives_array_val = train_model(model, alpha, n_iters,
x_train_gpu[:, None, :, :], y_train_gpu,
x_val_gpu[:, None, :, :], y_val_gpu,
llambda)
## Plot the objective
## ==================
fig, ax = plt.subplots()
ax.plot(objectives_array_train, label='Train')
ax.plot(objectives_array_val, label='Validation')
ax.set_title('Otimization objective')
ax.set_xlabel('Step')
ax.set_xlabel('Objective')
ax.set_ylim(0, 5)
ax.legend();
with torch.no_grad():
y_hat_test = model(x_test_gpu[:, None, :, :]).argmax(dim=1)[:, 0, 0]
empirical_risk_test = (y_hat_test != y_test_gpu).float().mean()
print('The empirical risk (amount of misclassifications) on the test set is: {}'.format(empirical_risk_test))
###Output
_____no_output_____ |
stimuli/preprocess_neural_speaker_for_eval.ipynb | ###Markdown
helper funcs
###Code
## this helps to sort in human order
import re
def tryint(s):
try:
return int(s)
except ValueError:
return s
def alphanum_key(s):
""" Turn a string into a list of string and number chunks.
"z23a" -> ["z", 23, "a"]
"""
return [ tryint(c) for c in re.split('([0-9]+)', s) ]
def sort_nicely(l):
""" Sort the given list in the way that humans expect.
"""
l.sort(key=alphanum_key)
def load_text(path):
with open(path, 'r') as f:
x = f.readlines()
utt = x[0]
# replace special tokens with question marks
if '<DIA>' in utt:
utt = utt.replace('<DIA>', '-')
if '<UKN>' in utt:
utt = utt.replace('<UKN>', '___')
return utt
###Output
_____no_output_____
###Markdown
setup
###Code
# paths
alphanum = dict(zip(range(26),string.ascii_lowercase))
conditions = ['literal','pragmatic']
upload_dir = './context_agnostic_False_rs_33'
bucket_name = 'shapenet-chairs-speaker-eval'
dataset_name = 'shapenet_chairs_speaker_eval'
# get list of triplet dirs
triplet_dirs = [i for i in os.listdir(upload_dir) if i != '.DS_Store']
triplet_dirs = [i for i in triplet_dirs if i[:7]=='triplet']
triplet_dirs = [os.path.join(upload_dir,i) for i in triplet_dirs]
sort_nicely(triplet_dirs)
# go through and rename the images from 0,1,2 to distractor1,distractor2,target
for this_triplet in triplet_dirs:
if os.path.exists(os.path.join(this_triplet,'0.png')):
_shapenet_ids = np.load(os.path.join(this_triplet,'shape_net_ids.npy'))
shapenet_id_dict = dict(zip(['distractor1','distractor2','target'],_shapenet_ids))
os.rename(os.path.join(this_triplet,'0.png'),os.path.join(this_triplet,'{}_distractor1.png'.format(shapenet_id_dict['distractor1'])))
os.rename(os.path.join(this_triplet,'1.png'),os.path.join(this_triplet,'{}_distractor2.png'.format(shapenet_id_dict['distractor2'])))
os.rename(os.path.join(this_triplet,'2.png'),os.path.join(this_triplet,'{}_target.png'.format(shapenet_id_dict['target'])))
# _shapenet_ids = np.load(os.path.join(this_triplet,'shape_net_ids.npy'))
# shapenet_id_dict = dict(zip(['distractor1','distractor2','target'],_shapenet_ids))
# literal_utt = load_text(os.path.join(this_triplet,'literal_utterance.txt'))
# pragmatic_utt = load_text(os.path.join(this_triplet,'pragmatic_utterance.txt'))
###Output
_____no_output_____
###Markdown
upload stims to s3
###Code
import boto
runThis = 0
if runThis:
conn = boto.connect_s3()
b = conn.create_bucket(bucket_name) ### if bucket already exists, then get_bucket, else create_bucket
for ind,this_triplet in enumerate(triplet_dirs):
ims = [i for i in os.listdir(this_triplet) if i[-3:]=='png']
for im in ims:
print ind, im
k = b.new_key(im)
k.set_contents_from_filename(os.path.join(this_triplet,im))
k.set_acl('public-read')
###Output
_____no_output_____
###Markdown
build stimulus dictionary & upload metadata to mongo
###Code
print('Generating list of triplets and their attributes...')
# generate pandas dataframe with different attributes
condition = []
family = []
utt = []
target = []
distractor1 = []
distractor2 = []
games = [] # this field keeps track of which games this triplet has been shown in
shuffler_ind = []
## generate permuted list of triplet indices in order to be able retrieve from triplets pseudorandomly
inds = np.arange(len(conditions)*len(triplet_dirs))
shuffled_inds = np.random.RandomState(0).permutation(inds)
counter = 0
for cond_ind,this_condition in enumerate(conditions):
for trip_ind,this_triplet in enumerate(triplet_dirs):
ims = [i for i in os.listdir(this_triplet) if i[-3:]=='png']
# extract filename
target_filename = [i for i in ims if 'target' in i][0]
distractor1_filename = [i for i in ims if 'distractor1' in i][0]
distractor2_filename = [i for i in ims if 'distractor2' in i][0]
# define url
target_url = 'https://s3.amazonaws.com/{}/{}'.format(bucket_name,target_filename)
distractor1_url = 'https://s3.amazonaws.com/{}/{}'.format(bucket_name,distractor1_filename)
distractor2_url = 'https://s3.amazonaws.com/{}/{}'.format(bucket_name,distractor2_filename)
# extract shapenetid
target_shapenetid = target_filename.split('_')[0]
distractor1_shapenetid = distractor1_filename.split('_')[0]
distractor2_shapenetid = distractor2_filename.split('_')[0]
# roll metadata into targ, d1, d2 dictionaries
_target = {'filename': target_filename, 'url': target_url, 'shapenetid': target_shapenetid}
_distractor1 = {'filename': distractor1_filename, 'url': distractor1_url, 'shapenetid': distractor1_shapenetid}
_distractor2 = {'filename': distractor2_filename, 'url': distractor2_url, 'shapenetid': distractor2_shapenetid}
# extract family and utt info
this_family = this_triplet.split('/')[-1]
this_utt = load_text(os.path.join(this_triplet,'{}_utterance.txt'.format(this_condition)))
# append to lists to prep for dataframe
condition.append(this_condition)
family.append(this_family)
utt.append(this_utt)
target.append(_target)
distractor1.append(_distractor1)
distractor2.append(_distractor2)
games.append([])
shuffler_ind.append(shuffled_inds[counter])
counter += 1
print('Generating pandas dataframe...')
table = [condition,family,utt,target,distractor1,distractor2,games,shuffler_ind]
headers = ['condition','family','utt','target','distractor1','distractor2','games','shuffler_ind']
df = pd.DataFrame(table)
df = df.transpose()
df.columns = headers
## save out to file
print('Saving out json dictionary out to file...')
stimdict = df.to_dict(orient='records')
with open('{}.js'.format(dataset_name), 'w') as fout:
json.dump(stimdict, fout)
### next todo is to upload this JSON to initialize the new stimulus collection
print('next todo is to upload this JSON to initialize the new stimulus collection...')
import json
J = json.loads(open('{}.js'.format(dataset_name),mode='ru').read())
##assert len(J)==len(all_files)
print 'dataset_name: {}'.format(dataset_name)
print len(J)
# set vars
auth = pd.read_csv('auth.txt', header = None) # this auth.txt file contains the password for the sketchloop user
pswd = auth.values[0][0]
user = 'sketchloop'
host = 'rxdhawkins.me' ## cocolab ip address
# have to fix this to be able to analyze from local
conn = pm.MongoClient('mongodb://sketchloop:' + pswd + '@127.0.0.1')
db = conn['stimuli']
coll = db[dataset_name]
## actually add data now to the database
reallyRun = 1
if reallyRun:
for (i,j) in enumerate(J):
if i%100==0:
print ('%d of %d' % (i,len(J)))
coll.insert_one(j)
## check how many records have been retrieved
a = coll.find({'shuffler_ind':{'$gte':0}})
numGames = []
for rec in a:
numGames.append(len(rec['games']))
b = np.array(numGames)
print np.mean(b>0)
###Output
_____no_output_____ |
notebooks/CLEU-2020.ipynb | ###Markdown
IOS-XE CLEU 2020 Demo Connecting to a Device Let's define some variables:
###Code
# Local CSR 1000v (running under vagrant) -- rtr1
HOST = '127.0.0.1'
PORT = 2223
USER = 'vagrant'
PASS = 'vagrant'
###Output
_____no_output_____
###Markdown
Now let's establish a NETCONF session to that box using ncclient:
###Code
from ncclient import manager
from lxml import etree
def pretty_print(retval):
print(etree.tostring(retval.data, pretty_print=True).decode())
def my_unknown_host_cb(host, fingerprint):
return True
m = manager.connect(host=HOST, port=PORT, username=USER, password=PASS,
allow_agent=False,
look_for_keys=False,
hostkey_verify=False,
unknown_host_cb=my_unknown_host_cb)
m
###Output
_____no_output_____
###Markdown
Capabilities Let's look at the capabilities presented by the thing we've just connected to:
###Code
for c in m.server_capabilities:
print(c)
###Output
urn:ietf:params:netconf:base:1.0
urn:ietf:params:netconf:base:1.1
urn:ietf:params:netconf:capability:writable-running:1.0
urn:ietf:params:netconf:capability:xpath:1.0
urn:ietf:params:netconf:capability:validate:1.0
urn:ietf:params:netconf:capability:validate:1.1
urn:ietf:params:netconf:capability:rollback-on-error:1.0
urn:ietf:params:netconf:capability:notification:1.0
urn:ietf:params:netconf:capability:interleave:1.0
urn:ietf:params:netconf:capability:with-defaults:1.0?basic-mode=explicit&also-supported=report-all-tagged
urn:ietf:params:netconf:capability:yang-library:1.0?revision=2016-06-21&module-set-id=61de1a2313e60afe62df413a77ed9087
http://tail-f.com/ns/netconf/actions/1.0
http://tail-f.com/ns/netconf/extensions
http://cisco.com/ns/cisco-xe-ietf-ip-deviation?module=cisco-xe-ietf-ip-deviation&revision=2016-08-10
http://cisco.com/ns/cisco-xe-ietf-ipv4-unicast-routing-deviation?module=cisco-xe-ietf-ipv4-unicast-routing-deviation&revision=2015-09-11
http://cisco.com/ns/cisco-xe-ietf-ipv6-unicast-routing-deviation?module=cisco-xe-ietf-ipv6-unicast-routing-deviation&revision=2015-09-11
http://cisco.com/ns/cisco-xe-ietf-ospf-deviation?module=cisco-xe-ietf-ospf-deviation&revision=2018-02-09
http://cisco.com/ns/cisco-xe-ietf-routing-deviation?module=cisco-xe-ietf-routing-deviation&revision=2016-07-09
http://cisco.com/ns/cisco-xe-openconfig-acl-deviation?module=cisco-xe-openconfig-acl-deviation&revision=2017-08-25
http://cisco.com/ns/cisco-xe-openconfig-aft-deviation?module=cisco-xe-openconfig-aft-deviation&revision=2018-12-05
http://cisco.com/ns/cisco-xe-openconfig-isis-deviation?module=cisco-xe-openconfig-isis-deviation&revision=2018-12-05
http://cisco.com/ns/cisco-xe-openconfig-lldp-deviation?module=cisco-xe-openconfig-lldp-deviation&revision=2018-07-25
http://cisco.com/ns/cisco-xe-openconfig-mpls-deviation?module=cisco-xe-openconfig-mpls-deviation&revision=2019-06-27
http://cisco.com/ns/cisco-xe-openconfig-segment-routing-deviation?module=cisco-xe-openconfig-segment-routing-deviation&revision=2018-12-05
http://cisco.com/ns/cisco-xe-routing-openconfig-system-management-deviation?module=cisco-xe-routing-openconfig-system-management-deviation&revision=2019-07-01
http://cisco.com/ns/mpls-static/devs?module=common-mpls-static-devs&revision=2015-09-11
http://cisco.com/ns/nvo/devs?module=nvo-devs&revision=2015-09-11
http://cisco.com/ns/yang/Cisco-IOS-XE-aaa?module=Cisco-IOS-XE-aaa&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-aaa-oper?module=Cisco-IOS-XE-aaa-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-acl?module=Cisco-IOS-XE-acl&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-acl-oper?module=Cisco-IOS-XE-acl-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-app-hosting-cfg?module=Cisco-IOS-XE-app-hosting-cfg&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-app-hosting-oper?module=Cisco-IOS-XE-app-hosting-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-arp?module=Cisco-IOS-XE-arp&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-arp-oper?module=Cisco-IOS-XE-arp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-atm?module=Cisco-IOS-XE-atm&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bba-group?module=Cisco-IOS-XE-bba-group&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bfd?module=Cisco-IOS-XE-bfd&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bfd-oper?module=Cisco-IOS-XE-bfd-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bgp?module=Cisco-IOS-XE-bgp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bgp-common-oper?module=Cisco-IOS-XE-bgp-common-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bgp-oper?module=Cisco-IOS-XE-bgp-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bgp-route-oper?module=Cisco-IOS-XE-bgp-route-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-bridge-domain?module=Cisco-IOS-XE-bridge-domain&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-call-home?module=Cisco-IOS-XE-call-home&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-card?module=Cisco-IOS-XE-card&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cdp?module=Cisco-IOS-XE-cdp&revision=2019-11-01&deviations=Cisco-IOS-XE-cdp-deviation
http://cisco.com/ns/yang/Cisco-IOS-XE-cdp-deviation?module=Cisco-IOS-XE-cdp-deviation&revision=2019-07-23
http://cisco.com/ns/yang/Cisco-IOS-XE-cdp-oper?module=Cisco-IOS-XE-cdp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cef?module=Cisco-IOS-XE-cef&revision=2019-11-01&features=asr1k-dpi
http://cisco.com/ns/yang/Cisco-IOS-XE-cellular?module=Cisco-IOS-XE-cellular&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cellular-rpc?module=Cisco-IOS-XE-cellular-rpc&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cellwan-oper?module=Cisco-IOS-XE-cellwan-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cfm-oper?module=Cisco-IOS-XE-cfm-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-checkpoint-archive-oper?module=Cisco-IOS-XE-checkpoint-archive-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-common-types?module=Cisco-IOS-XE-common-types&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-controller?module=Cisco-IOS-XE-controller&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-crypto?module=Cisco-IOS-XE-crypto&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-crypto-oper?module=Cisco-IOS-XE-crypto-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-crypto-pki-events?module=Cisco-IOS-XE-crypto-pki-events&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-crypto-pki-oper?module=Cisco-IOS-XE-crypto-pki-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cts?module=Cisco-IOS-XE-cts&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-cts-rpc?module=Cisco-IOS-XE-cts-rpc&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-dapr?module=Cisco-IOS-XE-dapr&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-device-hardware-oper?module=Cisco-IOS-XE-device-hardware-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-device-tracking?module=Cisco-IOS-XE-device-tracking&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-dhcp?module=Cisco-IOS-XE-dhcp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-dhcp-oper?module=Cisco-IOS-XE-dhcp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-diagnostics?module=Cisco-IOS-XE-diagnostics&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-dialer?module=Cisco-IOS-XE-dialer&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-dot1x?module=Cisco-IOS-XE-dot1x&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-eem?module=Cisco-IOS-XE-eem&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-efp-oper?module=Cisco-IOS-XE-efp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-eigrp?module=Cisco-IOS-XE-eigrp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-eigrp-oper?module=Cisco-IOS-XE-eigrp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-environment-oper?module=Cisco-IOS-XE-environment-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-eta?module=Cisco-IOS-XE-eta&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ethernet?module=Cisco-IOS-XE-ethernet&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-event-history-types?module=Cisco-IOS-XE-event-history-types&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ezpm?module=Cisco-IOS-XE-ezpm&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-features?module=Cisco-IOS-XE-features&revision=2019-10-01&features=virtual-template,routing-platform,punt-num,parameter-map,multilink,l2vpn,l2,ezpm,eth-evc,esmc,efp,dhcp-border-relay,crypto
http://cisco.com/ns/yang/Cisco-IOS-XE-fib-oper?module=Cisco-IOS-XE-fib-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-flow?module=Cisco-IOS-XE-flow&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-flow-monitor-oper?module=Cisco-IOS-XE-flow-monitor-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-flow-rpc?module=Cisco-IOS-XE-flow-rpc&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-fw-oper?module=Cisco-IOS-XE-fw-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-gir-oper?module=Cisco-IOS-XE-gir-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-http?module=Cisco-IOS-XE-http&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-icmp?module=Cisco-IOS-XE-icmp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-igmp?module=Cisco-IOS-XE-igmp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-im-events-oper?module=Cisco-IOS-XE-im-events-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-interface-common?module=Cisco-IOS-XE-interface-common&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper?module=Cisco-IOS-XE-interfaces-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ios-common-oper?module=Cisco-IOS-XE-ios-common-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ios-events-oper?module=Cisco-IOS-XE-ios-events-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ip-sla-oper?module=Cisco-IOS-XE-ip-sla-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ipv6-oper?module=Cisco-IOS-XE-ipv6-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-isis?module=Cisco-IOS-XE-isis&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-iwanfabric?module=Cisco-IOS-XE-iwanfabric&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-l2vpn?module=Cisco-IOS-XE-l2vpn&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-l3vpn?module=Cisco-IOS-XE-l3vpn&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-lisp?module=Cisco-IOS-XE-lisp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-lisp-oper?module=Cisco-IOS-XE-lisp-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-lldp?module=Cisco-IOS-XE-lldp&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-lldp-oper?module=Cisco-IOS-XE-lldp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mdns-gateway?module=Cisco-IOS-XE-mdns-gateway&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mdt-cfg?module=Cisco-IOS-XE-mdt-cfg&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mdt-common-defs?module=Cisco-IOS-XE-mdt-common-defs&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mdt-oper?module=Cisco-IOS-XE-mdt-oper&revision=2019-09-04
http://cisco.com/ns/yang/Cisco-IOS-XE-memory-oper?module=Cisco-IOS-XE-memory-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mld?module=Cisco-IOS-XE-mld&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mlppp-oper?module=Cisco-IOS-XE-mlppp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mpls?module=Cisco-IOS-XE-mpls&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-mpls-forwarding-oper?module=Cisco-IOS-XE-mpls-forwarding-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-multicast?module=Cisco-IOS-XE-multicast&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-nam?module=Cisco-IOS-XE-nam&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-nat?module=Cisco-IOS-XE-nat&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-nat-oper?module=Cisco-IOS-XE-nat-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-native?module=Cisco-IOS-XE-native&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-nbar?module=Cisco-IOS-XE-nbar&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-nd?module=Cisco-IOS-XE-nd&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-nhrp?module=Cisco-IOS-XE-nhrp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ntp?module=Cisco-IOS-XE-ntp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ntp-oper?module=Cisco-IOS-XE-ntp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-object-group?module=Cisco-IOS-XE-object-group&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ospf?module=Cisco-IOS-XE-ospf&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ospf-oper?module=Cisco-IOS-XE-ospf-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ospfv3?module=Cisco-IOS-XE-ospfv3&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-otv?module=Cisco-IOS-XE-otv&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-pathmgr?module=Cisco-IOS-XE-pathmgr&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-perf-measure?module=Cisco-IOS-XE-perf-measure&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-pfr?module=Cisco-IOS-XE-pfr&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-platform?module=Cisco-IOS-XE-platform&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-platform-oper?module=Cisco-IOS-XE-platform-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-platform-software-oper?module=Cisco-IOS-XE-platform-software-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-pnp?module=Cisco-IOS-XE-pnp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-policy?module=Cisco-IOS-XE-policy&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ppp?module=Cisco-IOS-XE-ppp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-ppp-oper?module=Cisco-IOS-XE-ppp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-pppoe?module=Cisco-IOS-XE-pppoe&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-process-cpu-oper?module=Cisco-IOS-XE-process-cpu-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-process-memory-oper?module=Cisco-IOS-XE-process-memory-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-qfp-stats?module=Cisco-IOS-XE-qfp-stats&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-qos?module=Cisco-IOS-XE-qos&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-rip?module=Cisco-IOS-XE-rip&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-rmi-dad?module=Cisco-IOS-XE-rmi-dad&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-route-map?module=Cisco-IOS-XE-route-map&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-rpc?module=Cisco-IOS-XE-rpc&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-rsvp?module=Cisco-IOS-XE-rsvp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-sanet?module=Cisco-IOS-XE-sanet&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-segment-routing?module=Cisco-IOS-XE-segment-routing&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-service-discovery?module=Cisco-IOS-XE-service-discovery&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-service-insertion?module=Cisco-IOS-XE-service-insertion&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-service-routing?module=Cisco-IOS-XE-service-routing&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-site-manager?module=Cisco-IOS-XE-site-manager&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-sla?module=Cisco-IOS-XE-sla&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-sm-enum-types?module=Cisco-IOS-XE-sm-enum-types&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-sm-events-oper?module=Cisco-IOS-XE-sm-events-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-snmp?module=Cisco-IOS-XE-snmp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-spanning-tree?module=Cisco-IOS-XE-spanning-tree&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-switch?module=Cisco-IOS-XE-switch&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-track?module=Cisco-IOS-XE-track&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-trustsec-oper?module=Cisco-IOS-XE-trustsec-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-tunnel?module=Cisco-IOS-XE-tunnel&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-types?module=Cisco-IOS-XE-types&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-umbrella?module=Cisco-IOS-XE-umbrella&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-umbrella-oper?module=Cisco-IOS-XE-umbrella-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-umbrella-oper-dp?module=Cisco-IOS-XE-umbrella-oper-dp&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-umbrella-rpc?module=Cisco-IOS-XE-umbrella-rpc&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-utd?module=Cisco-IOS-XE-utd&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-utd-common-oper?module=Cisco-IOS-XE-utd-common-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-utd-oper?module=Cisco-IOS-XE-utd-oper&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-utd-rpc?module=Cisco-IOS-XE-utd-rpc&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vlan?module=Cisco-IOS-XE-vlan&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-voice?module=Cisco-IOS-XE-voice&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vpdn?module=Cisco-IOS-XE-vpdn&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vrf-oper?module=Cisco-IOS-XE-vrf-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vrrp?module=Cisco-IOS-XE-vrrp&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vrrp-oper?module=Cisco-IOS-XE-vrrp-oper&revision=2019-05-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vservice?module=Cisco-IOS-XE-vservice&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vtp?module=Cisco-IOS-XE-vtp&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-vxlan?module=Cisco-IOS-XE-vxlan&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-wccp?module=Cisco-IOS-XE-wccp&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-wsma?module=Cisco-IOS-XE-wsma&revision=2019-07-01
http://cisco.com/ns/yang/Cisco-IOS-XE-yang-interfaces-cfg?module=Cisco-IOS-XE-yang-interfaces-cfg&revision=2019-05-21
http://cisco.com/ns/yang/Cisco-IOS-XE-zone?module=Cisco-IOS-XE-zone&revision=2019-11-01
http://cisco.com/ns/yang/Cisco-IOS-XE-zone-rpc?module=Cisco-IOS-XE-zone-rpc&revision=2019-11-01
http://cisco.com/ns/yang/cisco-semver?module=cisco-semver&revision=2019-03-20
http://cisco.com/ns/yang/cisco-smart-license?module=cisco-smart-license&revision=2019-07-01
http://cisco.com/ns/yang/cisco-xe-bgp-policy-deviation?module=cisco-xe-openconfig-bgp-policy-deviation&revision=2017-07-24
http://cisco.com/ns/yang/cisco-xe-ietf-event-notifications-deviation?module=cisco-xe-ietf-event-notifications-deviation&revision=2018-12-03
http://cisco.com/ns/yang/cisco-xe-ietf-yang-push-deviation?module=cisco-xe-ietf-yang-push-deviation&revision=2018-12-03
http://cisco.com/ns/yang/cisco-xe-openconfig-acl-ext?module=cisco-xe-openconfig-acl-ext&revision=2017-03-30
http://cisco.com/ns/yang/cisco-xe-openconfig-bgp-deviation?module=cisco-xe-openconfig-bgp-deviation&revision=2018-05-21
http://cisco.com/ns/yang/cisco-xe-openconfig-if-ethernet-ext?module=cisco-xe-openconfig-if-ethernet-ext&revision=2017-10-30
http://cisco.com/ns/yang/cisco-xe-openconfig-interfaces-ext?module=cisco-xe-openconfig-interfaces-ext&revision=2018-07-14
http://cisco.com/ns/yang/cisco-xe-openconfig-network-instance-deviation?module=cisco-xe-openconfig-network-instance-deviation&revision=2017-02-14
http://cisco.com/ns/yang/cisco-xe-openconfig-rib-bgp-ext?module=cisco-xe-openconfig-rib-bgp-ext&revision=2016-11-30
http://cisco.com/ns/yang/cisco-xe-openconfig-system-ext?module=cisco-xe-openconfig-system-ext&revision=2018-03-21
http://cisco.com/ns/yang/cisco-xe-routing-openconfig-vlan-deviation?module=cisco-xe-routing-openconfig-vlan-deviation&revision=2018-12-12
http://cisco.com/ns/yang/cisco-xe-routing-policy-deviation?module=cisco-xe-openconfig-routing-policy-deviation&revision=2017-03-30
http://cisco.com/ns/yang/ios-xe/template?module=Cisco-IOS-XE-template&revision=2019-11-01
http://cisco.com/yang/cisco-ia?module=cisco-ia&revision=2019-07-01
http://cisco.com/yang/cisco-self-mgmt?module=cisco-self-mgmt&revision=2019-07-01
http://openconfig.net/yang/aaa?module=openconfig-aaa&revision=2017-09-18
http://openconfig.net/yang/aaa/types?module=openconfig-aaa-types&revision=2017-09-18
http://openconfig.net/yang/acl?module=openconfig-acl&revision=2017-05-26&deviations=cisco-xe-openconfig-acl-deviation
http://openconfig.net/yang/aft?module=openconfig-aft&revision=2017-01-13
http://openconfig.net/yang/aft/ni?module=openconfig-aft-network-instance&revision=2017-01-13
http://openconfig.net/yang/alarms?module=openconfig-alarms&revision=2017-08-24
http://openconfig.net/yang/alarms/types?module=openconfig-alarm-types&revision=2018-11-21
http://openconfig.net/yang/bgp?module=openconfig-bgp&revision=2016-06-21
http://openconfig.net/yang/bgp-policy?module=openconfig-bgp-policy&revision=2016-06-21&deviations=cisco-xe-openconfig-bgp-policy-deviation
http://openconfig.net/yang/bgp-types?module=openconfig-bgp-types&revision=2016-06-21
http://openconfig.net/yang/cisco-xe-openconfig-if-ip-deviation?module=cisco-xe-openconfig-if-ip-deviation&revision=2017-03-04
http://openconfig.net/yang/cisco-xe-openconfig-interfaces-deviation?module=cisco-xe-openconfig-interfaces-deviation&revision=2018-08-21
http://openconfig.net/yang/cisco-xe-routing-csr-openconfig-platform-deviation?module=cisco-xe-routing-csr-openconfig-platform-deviation&revision=2010-10-09
http://openconfig.net/yang/cisco-xe-routing-openconfig-system-deviation?module=cisco-xe-routing-openconfig-system-deviation&revision=2017-11-27
http://openconfig.net/yang/fib-types?module=openconfig-aft-types&revision=2017-01-13
http://openconfig.net/yang/header-fields?module=openconfig-packet-match&revision=2017-05-26
http://openconfig.net/yang/interfaces?module=openconfig-interfaces&revision=2018-01-05&deviations=cisco-xe-openconfig-interfaces-deviation
http://openconfig.net/yang/interfaces/aggregate?module=openconfig-if-aggregate&revision=2018-01-05
http://openconfig.net/yang/interfaces/ethernet?module=openconfig-if-ethernet&revision=2018-01-05
http://openconfig.net/yang/interfaces/ip?module=openconfig-if-ip&revision=2018-01-05&deviations=cisco-xe-openconfig-if-ip-deviation,cisco-xe-openconfig-interfaces-deviation
http://openconfig.net/yang/interfaces/ip-ext?module=openconfig-if-ip-ext&revision=2018-01-05
http://openconfig.net/yang/isis-lsdb-types?module=openconfig-isis-lsdb-types&revision=2017-01-13
http://openconfig.net/yang/isis-types?module=openconfig-isis-types&revision=2017-01-13
http://openconfig.net/yang/lacp?module=openconfig-lacp&revision=2016-05-26
http://openconfig.net/yang/ldp?module=openconfig-mpls-ldp&revision=2016-12-15
http://openconfig.net/yang/lldp?module=openconfig-lldp&revision=2016-05-16&deviations=cisco-xe-openconfig-lldp-deviation
http://openconfig.net/yang/lldp/types?module=openconfig-lldp-types&revision=2016-05-16
http://openconfig.net/yang/local-routing?module=openconfig-local-routing&revision=2016-05-11
http://openconfig.net/yang/mpls?module=openconfig-mpls&revision=2016-12-15&deviations=cisco-xe-openconfig-mpls-deviation
http://openconfig.net/yang/mpls-sr?module=openconfig-mpls-sr&revision=2016-12-15
http://openconfig.net/yang/mpls-types?module=openconfig-mpls-types&revision=2016-12-15
http://openconfig.net/yang/network-instance?module=openconfig-network-instance&revision=2017-01-13&deviations=cisco-xe-openconfig-aft-deviation,cisco-xe-openconfig-bgp-deviation,cisco-xe-openconfig-isis-deviation,cisco-xe-openconfig-mpls-deviation,cisco-xe-openconfig-network-instance-deviation,cisco-xe-openconfig-segment-routing-deviation
http://openconfig.net/yang/network-instance-l3?module=openconfig-network-instance-l3&revision=2017-01-13
http://openconfig.net/yang/network-instance-types?module=openconfig-network-instance-types&revision=2016-12-15
http://openconfig.net/yang/openconfig-ext?module=openconfig-extensions&revision=2018-10-17
http://openconfig.net/yang/openconfig-isis?module=openconfig-isis&revision=2017-01-13
http://openconfig.net/yang/openconfig-isis-policy?module=openconfig-isis-policy&revision=2017-01-13
http://openconfig.net/yang/openconfig-types?module=openconfig-types&revision=2018-11-21
http://openconfig.net/yang/packet-match-types?module=openconfig-packet-match-types&revision=2017-05-26
http://openconfig.net/yang/platform?module=openconfig-platform&revision=2018-11-21&deviations=cisco-xe-routing-csr-openconfig-platform-deviation
http://openconfig.net/yang/platform-types?module=openconfig-platform-types&revision=2018-11-21
http://openconfig.net/yang/policy-types?module=openconfig-policy-types&revision=2016-05-12
http://openconfig.net/yang/rib/bgp?module=openconfig-rib-bgp&revision=2017-03-07
http://openconfig.net/yang/rib/bgp-ext?module=openconfig-rib-bgp-ext&revision=2016-04-11
http://openconfig.net/yang/rib/bgp-types?module=openconfig-rib-bgp-types&revision=2016-04-11
http://openconfig.net/yang/routing-policy?module=openconfig-routing-policy&revision=2016-05-12&deviations=cisco-xe-openconfig-routing-policy-deviation
http://openconfig.net/yang/rsvp?module=openconfig-mpls-rsvp&revision=2016-12-15
http://openconfig.net/yang/sr?module=openconfig-segment-routing&revision=2017-01-12
http://openconfig.net/yang/system?module=openconfig-system&revision=2018-07-17&deviations=cisco-xe-routing-openconfig-system-deviation,cisco-xe-routing-openconfig-system-management-deviation
http://openconfig.net/yang/system/logging?module=openconfig-system-logging&revision=2017-09-18
http://openconfig.net/yang/system/management?module=openconfig-system-management&revision=2018-11-21
http://openconfig.net/yang/system/procmon?module=openconfig-procmon&revision=2017-09-18
http://openconfig.net/yang/system/terminal?module=openconfig-system-terminal&revision=2017-09-18
http://openconfig.net/yang/types/inet?module=openconfig-inet-types&revision=2017-08-24
http://openconfig.net/yang/types/yang?module=openconfig-yang-types&revision=2018-11-21
http://openconfig.net/yang/vlan?module=openconfig-vlan&revision=2016-05-26&deviations=cisco-xe-routing-openconfig-vlan-deviation
http://openconfig.net/yang/vlan-types?module=openconfig-vlan-types&revision=2016-05-26
http://tail-f.com/ns/common/query?module=tailf-common-query&revision=2017-12-15
http://tail-f.com/yang/common?module=tailf-common&revision=2019-03-18
http://tail-f.com/yang/common-monitoring?module=tailf-common-monitoring&revision=2013-06-14
http://tail-f.com/yang/confd-monitoring?module=tailf-confd-monitoring&revision=2013-06-14
http://tail-f.com/yang/netconf-monitoring?module=tailf-netconf-monitoring&revision=2019-03-28
urn:cisco:params:xml:ns:yang:cisco-bridge-common?module=cisco-bridge-common&revision=2019-07-01&features=configurable-bd-mac-limit-notif,configurable-bd-mac-limit-max,configurable-bd-mac-limit-actions,configurable-bd-mac-aging-types,configurable-bd-flooding-control
urn:cisco:params:xml:ns:yang:cisco-bridge-domain?module=cisco-bridge-domain&revision=2019-07-01&features=parameterized-bridge-domains,configurable-bd-storm-control,configurable-bd-static-mac,configurable-bd-snooping-profiles,configurable-bd-sh-group-number,configurable-bd-mtu,configurable-bd-member-features,configurable-bd-mac-secure,configurable-bd-mac-features,configurable-bd-mac-event-action,configurable-bd-ipsg,configurable-bd-groups,configurable-bd-flooding-mode,configurable-bd-flooding,configurable-bd-dai,clear-bridge-domain
urn:cisco:params:xml:ns:yang:cisco-ethernet?module=cisco-ethernet&revision=2016-05-10
urn:cisco:params:xml:ns:yang:cisco-routing-ext?module=cisco-routing-ext&revision=2019-11-01
urn:cisco:params:xml:ns:yang:cisco-storm-control?module=cisco-storm-control&revision=2019-07-01&features=configurable-storm-control-actions
urn:cisco:params:xml:ns:yang:cisco-xe-ietf-yang-push-ext?module=cisco-xe-ietf-yang-push-ext&revision=2019-03-26
urn:cisco:params:xml:ns:yang:pim?module=pim&revision=2019-07-01&features=bsr,auto-rp
urn:cisco:params:xml:ns:yang:pw?module=cisco-pw&revision=2019-07-01&features=static-label-direct-config,pw-vccv,pw-tag-impose-vlan-id,pw-status-config,pw-static-oam-config,pw-short-config,pw-sequencing,pw-preferred-path,pw-port-profiles,pw-oam-refresh-config,pw-mac-withdraw-config,pw-load-balancing,pw-ipv6-source,pw-interface,pw-grouping-config,pw-class-tag-rewrite,pw-class-switchover-delay,pw-class-status,pw-class-source-ip,pw-class-flow-setting,preferred-path-peer,predictive-redundancy-config,flow-label-tlv-code17,flow-label-static-config
urn:ietf:params:xml:ns:yang:c3pl-types?module=policy-types&revision=2019-07-01&features=protocol-name-support,match-wlan-user-priority-support,match-vpls-support,match-vlan-support,match-vlan-inner-support,match-src-mac-support,match-security-group-support,match-qos-group-support,match-prec-support,match-packet-length-support,match-mpls-exp-top-support,match-mpls-exp-imp-support,match-metadata-support,match-ipv6-acl-support,match-ipv6-acl-name-support,match-ipv4-acl-support,match-ipv4-acl-name-support,match-ip-rtp-support,match-input-interface-support,match-fr-dlci-support,match-fr-de-support,match-flow-record-support,match-flow-ip-support,match-dst-mac-support,match-discard-class-support,match-dei-support,match-dei-inner-support,match-cos-support,match-cos-inner-support,match-class-map-support,match-atm-vci-support,match-atm-clp-support,match-application-support
urn:ietf:params:xml:ns:yang:cisco-ospf?module=cisco-ospf&revision=2019-07-01&features=graceful-shutdown,flood-reduction,database-filter
urn:ietf:params:xml:ns:yang:cisco-policy?module=cisco-policy&revision=2019-07-01
urn:ietf:params:xml:ns:yang:cisco-policy-filters?module=cisco-policy-filters&revision=2019-07-01
urn:ietf:params:xml:ns:yang:cisco-policy-target?module=cisco-policy-target&revision=2019-07-01
urn:ietf:params:xml:ns:yang:common-mpls-static?module=common-mpls-static&revision=2019-07-01&deviations=common-mpls-static-devs
urn:ietf:params:xml:ns:yang:common-mpls-types?module=common-mpls-types&revision=2019-07-01
urn:ietf:params:xml:ns:yang:iana-crypt-hash?module=iana-crypt-hash&revision=2014-08-06&features=crypt-hash-sha-512,crypt-hash-sha-256,crypt-hash-md5
urn:ietf:params:xml:ns:yang:iana-if-type?module=iana-if-type&revision=2014-05-08
urn:ietf:params:xml:ns:yang:ietf-diffserv-action?module=ietf-diffserv-action&revision=2015-04-07&features=priority-rate-burst-support,hierarchial-policy-support,aqm-red-support
urn:ietf:params:xml:ns:yang:ietf-diffserv-classifier?module=ietf-diffserv-classifier&revision=2015-04-07&features=policy-inline-classifier-config
urn:ietf:params:xml:ns:yang:ietf-diffserv-policy?module=ietf-diffserv-policy&revision=2015-04-07&features=policy-template-support,hierarchial-policy-support
urn:ietf:params:xml:ns:yang:ietf-diffserv-target?module=ietf-diffserv-target&revision=2015-04-07&features=target-inline-policy-config
urn:ietf:params:xml:ns:yang:ietf-event-notifications?module=ietf-event-notifications&revision=2016-10-27&features=json,configured-subscriptions&deviations=cisco-xe-ietf-event-notifications-deviation,cisco-xe-ietf-yang-push-deviation
urn:ietf:params:xml:ns:yang:ietf-inet-types?module=ietf-inet-types&revision=2013-07-15
urn:ietf:params:xml:ns:yang:ietf-interfaces?module=ietf-interfaces&revision=2014-05-08&features=pre-provisioning,if-mib,arbitrary-names
urn:ietf:params:xml:ns:yang:ietf-interfaces-ext?module=ietf-interfaces-ext
urn:ietf:params:xml:ns:yang:ietf-ip?module=ietf-ip&revision=2014-06-16&features=ipv6-privacy-autoconf,ipv4-non-contiguous-netmasks&deviations=cisco-xe-ietf-ip-deviation
urn:ietf:params:xml:ns:yang:ietf-ipv4-unicast-routing?module=ietf-ipv4-unicast-routing&revision=2015-05-25&deviations=cisco-xe-ietf-ipv4-unicast-routing-deviation
urn:ietf:params:xml:ns:yang:ietf-ipv6-unicast-routing?module=ietf-ipv6-unicast-routing&revision=2015-05-25&deviations=cisco-xe-ietf-ipv6-unicast-routing-deviation
urn:ietf:params:xml:ns:yang:ietf-key-chain?module=ietf-key-chain&revision=2015-02-24&features=independent-send-accept-lifetime,hex-key-string,accept-tolerance
urn:ietf:params:xml:ns:yang:ietf-netconf-acm?module=ietf-netconf-acm&revision=2012-02-22
urn:ietf:params:xml:ns:yang:ietf-netconf-monitoring?module=ietf-netconf-monitoring&revision=2010-10-04
urn:ietf:params:xml:ns:yang:ietf-netconf-notifications?module=ietf-netconf-notifications&revision=2012-02-06
urn:ietf:params:xml:ns:yang:ietf-ospf?module=ietf-ospf&revision=2015-03-09&features=ttl-security,te-rid,router-id,remote-lfa,prefix-suppression,ospfv3-authentication-ipsec,nsr,node-flag,multi-topology,multi-area-adj,mtu-ignore,max-lsa,max-ecmp,lls,lfa,ldp-igp-sync,ldp-igp-autoconfig,interface-inheritance,instance-inheritance,graceful-restart,fast-reroute,demand-circuit,bfd,auto-cost,area-inheritance,admin-control&deviations=cisco-xe-ietf-ospf-deviation
urn:ietf:params:xml:ns:yang:ietf-restconf-monitoring?module=ietf-restconf-monitoring&revision=2017-01-26
urn:ietf:params:xml:ns:yang:ietf-routing?module=ietf-routing&revision=2015-05-25&features=router-id,multiple-ribs&deviations=cisco-xe-ietf-routing-deviation
urn:ietf:params:xml:ns:yang:ietf-yang-library?module=ietf-yang-library&revision=2016-06-21
urn:ietf:params:xml:ns:yang:ietf-yang-push?module=ietf-yang-push&revision=2016-10-28&features=on-change&deviations=cisco-xe-ietf-yang-push-deviation
urn:ietf:params:xml:ns:yang:ietf-yang-smiv2?module=ietf-yang-smiv2&revision=2012-06-22
urn:ietf:params:xml:ns:yang:ietf-yang-types?module=ietf-yang-types&revision=2013-07-15
urn:ietf:params:xml:ns:yang:nvo?module=nvo&revision=2019-07-01&deviations=nvo-devs
urn:ietf:params:xml:ns:yang:policy-attr?module=policy-attr&revision=2019-07-01
urn:ietf:params:xml:ns:yang:smiv2:ATM-FORUM-TC-MIB?module=ATM-FORUM-TC-MIB
urn:ietf:params:xml:ns:yang:smiv2:ATM-MIB?module=ATM-MIB&revision=1998-10-19
urn:ietf:params:xml:ns:yang:smiv2:ATM-TC-MIB?module=ATM-TC-MIB&revision=1998-10-19
urn:ietf:params:xml:ns:yang:smiv2:BGP4-MIB?module=BGP4-MIB&revision=1994-05-05
urn:ietf:params:xml:ns:yang:smiv2:BRIDGE-MIB?module=BRIDGE-MIB&revision=2005-09-19
urn:ietf:params:xml:ns:yang:smiv2:CISCO-AAA-SERVER-MIB?module=CISCO-AAA-SERVER-MIB&revision=2003-11-17
urn:ietf:params:xml:ns:yang:smiv2:CISCO-AAA-SESSION-MIB?module=CISCO-AAA-SESSION-MIB&revision=2006-03-21
urn:ietf:params:xml:ns:yang:smiv2:CISCO-AAL5-MIB?module=CISCO-AAL5-MIB&revision=2003-09-22
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ATM-EXT-MIB?module=CISCO-ATM-EXT-MIB&revision=2003-01-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ATM-PVCTRAP-EXTN-MIB?module=CISCO-ATM-PVCTRAP-EXTN-MIB&revision=2003-01-20
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ATM-QOS-MIB?module=CISCO-ATM-QOS-MIB&revision=2002-06-10
urn:ietf:params:xml:ns:yang:smiv2:CISCO-BGP-POLICY-ACCOUNTING-MIB?module=CISCO-BGP-POLICY-ACCOUNTING-MIB&revision=2002-07-26
urn:ietf:params:xml:ns:yang:smiv2:CISCO-BGP4-MIB?module=CISCO-BGP4-MIB&revision=2010-09-30
urn:ietf:params:xml:ns:yang:smiv2:CISCO-BULK-FILE-MIB?module=CISCO-BULK-FILE-MIB&revision=2002-06-10
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CBP-TARGET-MIB?module=CISCO-CBP-TARGET-MIB&revision=2006-05-24
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CBP-TARGET-TC-MIB?module=CISCO-CBP-TARGET-TC-MIB&revision=2006-03-24
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CBP-TC-MIB?module=CISCO-CBP-TC-MIB&revision=2008-06-24
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CDP-MIB?module=CISCO-CDP-MIB&revision=2005-03-21
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CEF-MIB?module=CISCO-CEF-MIB&revision=2006-01-30
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CEF-TC?module=CISCO-CEF-TC&revision=2005-09-30
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CONFIG-COPY-MIB?module=CISCO-CONFIG-COPY-MIB&revision=2005-04-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CONFIG-MAN-MIB?module=CISCO-CONFIG-MAN-MIB&revision=2007-04-27
urn:ietf:params:xml:ns:yang:smiv2:CISCO-CONTEXT-MAPPING-MIB?module=CISCO-CONTEXT-MAPPING-MIB&revision=2008-11-22
urn:ietf:params:xml:ns:yang:smiv2:CISCO-DATA-COLLECTION-MIB?module=CISCO-DATA-COLLECTION-MIB&revision=2002-10-30
urn:ietf:params:xml:ns:yang:smiv2:CISCO-DIAL-CONTROL-MIB?module=CISCO-DIAL-CONTROL-MIB&revision=2005-05-26
urn:ietf:params:xml:ns:yang:smiv2:CISCO-DOT3-OAM-MIB?module=CISCO-DOT3-OAM-MIB&revision=2006-05-31
urn:ietf:params:xml:ns:yang:smiv2:CISCO-DYNAMIC-TEMPLATE-MIB?module=CISCO-DYNAMIC-TEMPLATE-MIB&revision=2007-09-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-DYNAMIC-TEMPLATE-TC-MIB?module=CISCO-DYNAMIC-TEMPLATE-TC-MIB&revision=2012-01-27
urn:ietf:params:xml:ns:yang:smiv2:CISCO-EIGRP-MIB?module=CISCO-EIGRP-MIB&revision=2004-11-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-EMBEDDED-EVENT-MGR-MIB?module=CISCO-EMBEDDED-EVENT-MGR-MIB&revision=2006-11-07
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENHANCED-MEMPOOL-MIB?module=CISCO-ENHANCED-MEMPOOL-MIB&revision=2008-12-05
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENTITY-ALARM-MIB?module=CISCO-ENTITY-ALARM-MIB&revision=1999-07-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENTITY-EXT-MIB?module=CISCO-ENTITY-EXT-MIB&revision=2008-11-24
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENTITY-FRU-CONTROL-MIB?module=CISCO-ENTITY-FRU-CONTROL-MIB&revision=2013-08-19
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENTITY-QFP-MIB?module=CISCO-ENTITY-QFP-MIB&revision=2014-06-18
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENTITY-SENSOR-MIB?module=CISCO-ENTITY-SENSOR-MIB&revision=2015-01-15
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ENTITY-VENDORTYPE-OID-MIB?module=CISCO-ENTITY-VENDORTYPE-OID-MIB&revision=2014-12-09
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ETHER-CFM-MIB?module=CISCO-ETHER-CFM-MIB&revision=2004-12-28
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ETHERLIKE-EXT-MIB?module=CISCO-ETHERLIKE-EXT-MIB&revision=2010-06-04
urn:ietf:params:xml:ns:yang:smiv2:CISCO-FIREWALL-TC?module=CISCO-FIREWALL-TC&revision=2006-03-03
urn:ietf:params:xml:ns:yang:smiv2:CISCO-FLASH-MIB?module=CISCO-FLASH-MIB&revision=2013-08-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-FTP-CLIENT-MIB?module=CISCO-FTP-CLIENT-MIB&revision=2006-03-31
urn:ietf:params:xml:ns:yang:smiv2:CISCO-HSRP-EXT-MIB?module=CISCO-HSRP-EXT-MIB&revision=2010-09-02
urn:ietf:params:xml:ns:yang:smiv2:CISCO-HSRP-MIB?module=CISCO-HSRP-MIB&revision=2010-09-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-ATM2-PVCTRAP-MIB?module=CISCO-IETF-ATM2-PVCTRAP-MIB&revision=1998-02-03
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-ATM2-PVCTRAP-MIB-EXTN?module=CISCO-IETF-ATM2-PVCTRAP-MIB-EXTN&revision=2000-07-11
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-BFD-MIB?module=CISCO-IETF-BFD-MIB&revision=2011-04-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-FRR-MIB?module=CISCO-IETF-FRR-MIB&revision=2008-04-29
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-ISIS-MIB?module=CISCO-IETF-ISIS-MIB&revision=2005-08-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-MPLS-ID-STD-03-MIB?module=CISCO-IETF-MPLS-ID-STD-03-MIB&revision=2012-06-07
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-MPLS-TE-EXT-STD-03-MIB?module=CISCO-IETF-MPLS-TE-EXT-STD-03-MIB&revision=2012-06-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-PW-ATM-MIB?module=CISCO-IETF-PW-ATM-MIB&revision=2005-04-19
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-PW-ENET-MIB?module=CISCO-IETF-PW-ENET-MIB&revision=2002-09-22
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-PW-MIB?module=CISCO-IETF-PW-MIB&revision=2004-03-17
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-PW-MPLS-MIB?module=CISCO-IETF-PW-MPLS-MIB&revision=2003-02-26
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-PW-TC-MIB?module=CISCO-IETF-PW-TC-MIB&revision=2006-07-21
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IETF-PW-TDM-MIB?module=CISCO-IETF-PW-TDM-MIB&revision=2006-07-21
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IF-EXTENSION-MIB?module=CISCO-IF-EXTENSION-MIB&revision=2013-03-13
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IGMP-FILTER-MIB?module=CISCO-IGMP-FILTER-MIB&revision=2005-11-29
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IMAGE-LICENSE-MGMT-MIB?module=CISCO-IMAGE-LICENSE-MGMT-MIB&revision=2007-10-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IMAGE-MIB?module=CISCO-IMAGE-MIB&revision=1995-08-15
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IP-LOCAL-POOL-MIB?module=CISCO-IP-LOCAL-POOL-MIB&revision=2007-11-12
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IP-TAP-MIB?module=CISCO-IP-TAP-MIB&revision=2004-03-11
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IP-URPF-MIB?module=CISCO-IP-URPF-MIB&revision=2011-12-29
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPMROUTE-MIB?module=CISCO-IPMROUTE-MIB&revision=2005-03-07
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSEC-FLOW-MONITOR-MIB?module=CISCO-IPSEC-FLOW-MONITOR-MIB&revision=2007-10-24
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSEC-MIB?module=CISCO-IPSEC-MIB&revision=2000-08-07
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSEC-POLICY-MAP-MIB?module=CISCO-IPSEC-POLICY-MAP-MIB&revision=2000-08-17
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSLA-AUTOMEASURE-MIB?module=CISCO-IPSLA-AUTOMEASURE-MIB&revision=2007-06-13
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSLA-ECHO-MIB?module=CISCO-IPSLA-ECHO-MIB&revision=2007-08-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSLA-JITTER-MIB?module=CISCO-IPSLA-JITTER-MIB&revision=2007-07-24
urn:ietf:params:xml:ns:yang:smiv2:CISCO-IPSLA-TC-MIB?module=CISCO-IPSLA-TC-MIB&revision=2007-03-23
urn:ietf:params:xml:ns:yang:smiv2:CISCO-LICENSE-MGMT-MIB?module=CISCO-LICENSE-MGMT-MIB&revision=2012-04-19
urn:ietf:params:xml:ns:yang:smiv2:CISCO-MEDIA-GATEWAY-MIB?module=CISCO-MEDIA-GATEWAY-MIB&revision=2009-02-25
urn:ietf:params:xml:ns:yang:smiv2:CISCO-MPLS-LSR-EXT-STD-MIB?module=CISCO-MPLS-LSR-EXT-STD-MIB&revision=2012-04-30
urn:ietf:params:xml:ns:yang:smiv2:CISCO-MPLS-TC-EXT-STD-MIB?module=CISCO-MPLS-TC-EXT-STD-MIB&revision=2012-02-22
urn:ietf:params:xml:ns:yang:smiv2:CISCO-NBAR-PROTOCOL-DISCOVERY-MIB?module=CISCO-NBAR-PROTOCOL-DISCOVERY-MIB&revision=2002-08-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-NETSYNC-MIB?module=CISCO-NETSYNC-MIB&revision=2010-10-15
urn:ietf:params:xml:ns:yang:smiv2:CISCO-NTP-MIB?module=CISCO-NTP-MIB&revision=2006-07-31
urn:ietf:params:xml:ns:yang:smiv2:CISCO-OSPF-MIB?module=CISCO-OSPF-MIB&revision=2003-07-18
urn:ietf:params:xml:ns:yang:smiv2:CISCO-OSPF-TRAP-MIB?module=CISCO-OSPF-TRAP-MIB&revision=2003-07-18
urn:ietf:params:xml:ns:yang:smiv2:CISCO-PIM-MIB?module=CISCO-PIM-MIB&revision=2000-11-02
urn:ietf:params:xml:ns:yang:smiv2:CISCO-PING-MIB?module=CISCO-PING-MIB&revision=2001-08-28
urn:ietf:params:xml:ns:yang:smiv2:CISCO-PROCESS-MIB?module=CISCO-PROCESS-MIB&revision=2011-06-23
urn:ietf:params:xml:ns:yang:smiv2:CISCO-PRODUCTS-MIB?module=CISCO-PRODUCTS-MIB&revision=2014-11-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-PTP-MIB?module=CISCO-PTP-MIB&revision=2011-01-28
urn:ietf:params:xml:ns:yang:smiv2:CISCO-QOS-PIB-MIB?module=CISCO-QOS-PIB-MIB&revision=2007-08-29
urn:ietf:params:xml:ns:yang:smiv2:CISCO-RADIUS-EXT-MIB?module=CISCO-RADIUS-EXT-MIB&revision=2010-05-25
urn:ietf:params:xml:ns:yang:smiv2:CISCO-RF-MIB?module=CISCO-RF-MIB&revision=2005-09-01
urn:ietf:params:xml:ns:yang:smiv2:CISCO-RTTMON-MIB?module=CISCO-RTTMON-MIB&revision=2012-08-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-RTTMON-TC-MIB?module=CISCO-RTTMON-TC-MIB&revision=2012-05-25
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SESS-BORDER-CTRLR-CALL-STATS-MIB?module=CISCO-SESS-BORDER-CTRLR-CALL-STATS-MIB&revision=2010-09-03
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SESS-BORDER-CTRLR-STATS-MIB?module=CISCO-SESS-BORDER-CTRLR-STATS-MIB&revision=2010-09-15
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SIP-UA-MIB?module=CISCO-SIP-UA-MIB&revision=2004-02-19
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SMI?module=CISCO-SMI&revision=2012-08-29
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SONET-MIB?module=CISCO-SONET-MIB&revision=2003-03-07
urn:ietf:params:xml:ns:yang:smiv2:CISCO-ST-TC?module=CISCO-ST-TC&revision=2012-08-08
urn:ietf:params:xml:ns:yang:smiv2:CISCO-STP-EXTENSIONS-MIB?module=CISCO-STP-EXTENSIONS-MIB&revision=2013-03-07
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SUBSCRIBER-IDENTITY-TC-MIB?module=CISCO-SUBSCRIBER-IDENTITY-TC-MIB&revision=2011-12-23
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SUBSCRIBER-SESSION-MIB?module=CISCO-SUBSCRIBER-SESSION-MIB&revision=2012-08-08
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SUBSCRIBER-SESSION-TC-MIB?module=CISCO-SUBSCRIBER-SESSION-TC-MIB&revision=2012-01-27
urn:ietf:params:xml:ns:yang:smiv2:CISCO-SYSLOG-MIB?module=CISCO-SYSLOG-MIB&revision=2005-12-03
urn:ietf:params:xml:ns:yang:smiv2:CISCO-TAP2-MIB?module=CISCO-TAP2-MIB&revision=2009-11-06
urn:ietf:params:xml:ns:yang:smiv2:CISCO-TC?module=CISCO-TC&revision=2011-11-11
urn:ietf:params:xml:ns:yang:smiv2:CISCO-UBE-MIB?module=CISCO-UBE-MIB&revision=2010-11-29
urn:ietf:params:xml:ns:yang:smiv2:CISCO-UNIFIED-FIREWALL-MIB?module=CISCO-UNIFIED-FIREWALL-MIB&revision=2005-09-22
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VLAN-IFTABLE-RELATIONSHIP-MIB?module=CISCO-VLAN-IFTABLE-RELATIONSHIP-MIB&revision=2013-07-15
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VLAN-MEMBERSHIP-MIB?module=CISCO-VLAN-MEMBERSHIP-MIB&revision=2007-12-14
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VOICE-COMMON-DIAL-CONTROL-MIB?module=CISCO-VOICE-COMMON-DIAL-CONTROL-MIB&revision=2010-06-30
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VOICE-DIAL-CONTROL-MIB?module=CISCO-VOICE-DIAL-CONTROL-MIB&revision=2012-05-15
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VOICE-DNIS-MIB?module=CISCO-VOICE-DNIS-MIB&revision=2002-05-01
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VPDN-MGMT-MIB?module=CISCO-VPDN-MGMT-MIB&revision=2009-06-16
urn:ietf:params:xml:ns:yang:smiv2:CISCO-VTP-MIB?module=CISCO-VTP-MIB&revision=2013-10-14
urn:ietf:params:xml:ns:yang:smiv2:DIAL-CONTROL-MIB?module=DIAL-CONTROL-MIB&revision=1996-09-23
urn:ietf:params:xml:ns:yang:smiv2:DIFFSERV-DSCP-TC?module=DIFFSERV-DSCP-TC&revision=2002-05-09
urn:ietf:params:xml:ns:yang:smiv2:DIFFSERV-MIB?module=DIFFSERV-MIB&revision=2002-02-07
urn:ietf:params:xml:ns:yang:smiv2:DISMAN-EVENT-MIB?module=DISMAN-EVENT-MIB&revision=2000-10-16
urn:ietf:params:xml:ns:yang:smiv2:DISMAN-EXPRESSION-MIB?module=DISMAN-EXPRESSION-MIB&revision=2000-10-16
urn:ietf:params:xml:ns:yang:smiv2:DRAFT-MSDP-MIB?module=DRAFT-MSDP-MIB&revision=1999-12-16
urn:ietf:params:xml:ns:yang:smiv2:DS1-MIB?module=DS1-MIB&revision=1998-08-01
urn:ietf:params:xml:ns:yang:smiv2:DS3-MIB?module=DS3-MIB&revision=1998-08-01
urn:ietf:params:xml:ns:yang:smiv2:ENTITY-MIB?module=ENTITY-MIB&revision=2005-08-10
urn:ietf:params:xml:ns:yang:smiv2:ENTITY-SENSOR-MIB?module=ENTITY-SENSOR-MIB&revision=2002-12-16
urn:ietf:params:xml:ns:yang:smiv2:ENTITY-STATE-MIB?module=ENTITY-STATE-MIB&revision=2005-11-22
urn:ietf:params:xml:ns:yang:smiv2:ENTITY-STATE-TC-MIB?module=ENTITY-STATE-TC-MIB&revision=2005-11-22
urn:ietf:params:xml:ns:yang:smiv2:ETHER-WIS?module=ETHER-WIS&revision=2003-09-19
urn:ietf:params:xml:ns:yang:smiv2:EXPRESSION-MIB?module=EXPRESSION-MIB&revision=2005-11-24
urn:ietf:params:xml:ns:yang:smiv2:EtherLike-MIB?module=EtherLike-MIB&revision=2003-09-19
urn:ietf:params:xml:ns:yang:smiv2:FRAME-RELAY-DTE-MIB?module=FRAME-RELAY-DTE-MIB&revision=1997-05-01
urn:ietf:params:xml:ns:yang:smiv2:HCNUM-TC?module=HCNUM-TC&revision=2000-06-08
urn:ietf:params:xml:ns:yang:smiv2:IANA-ADDRESS-FAMILY-NUMBERS-MIB?module=IANA-ADDRESS-FAMILY-NUMBERS-MIB&revision=2000-09-08
urn:ietf:params:xml:ns:yang:smiv2:IANA-RTPROTO-MIB?module=IANA-RTPROTO-MIB&revision=2000-09-26
urn:ietf:params:xml:ns:yang:smiv2:IANAifType-MIB?module=IANAifType-MIB&revision=2006-03-31
urn:ietf:params:xml:ns:yang:smiv2:IEEE8021-TC-MIB?module=IEEE8021-TC-MIB&revision=2008-10-15
urn:ietf:params:xml:ns:yang:smiv2:IF-MIB?module=IF-MIB&revision=2000-06-14
urn:ietf:params:xml:ns:yang:smiv2:IGMP-STD-MIB?module=IGMP-STD-MIB&revision=2000-09-28
urn:ietf:params:xml:ns:yang:smiv2:INET-ADDRESS-MIB?module=INET-ADDRESS-MIB&revision=2005-02-04
urn:ietf:params:xml:ns:yang:smiv2:INT-SERV-MIB?module=INT-SERV-MIB&revision=1997-10-03
urn:ietf:params:xml:ns:yang:smiv2:INTEGRATED-SERVICES-MIB?module=INTEGRATED-SERVICES-MIB&revision=1995-11-03
urn:ietf:params:xml:ns:yang:smiv2:IP-FORWARD-MIB?module=IP-FORWARD-MIB&revision=1996-09-19
urn:ietf:params:xml:ns:yang:smiv2:IP-MIB?module=IP-MIB&revision=2006-02-02
urn:ietf:params:xml:ns:yang:smiv2:IPMROUTE-STD-MIB?module=IPMROUTE-STD-MIB&revision=2000-09-22
urn:ietf:params:xml:ns:yang:smiv2:IPV6-FLOW-LABEL-MIB?module=IPV6-FLOW-LABEL-MIB&revision=2003-08-28
urn:ietf:params:xml:ns:yang:smiv2:LLDP-MIB?module=LLDP-MIB&revision=2005-05-06
urn:ietf:params:xml:ns:yang:smiv2:MPLS-L3VPN-STD-MIB?module=MPLS-L3VPN-STD-MIB&revision=2006-01-23
urn:ietf:params:xml:ns:yang:smiv2:MPLS-LDP-GENERIC-STD-MIB?module=MPLS-LDP-GENERIC-STD-MIB&revision=2004-06-03
urn:ietf:params:xml:ns:yang:smiv2:MPLS-LDP-STD-MIB?module=MPLS-LDP-STD-MIB&revision=2004-06-03
urn:ietf:params:xml:ns:yang:smiv2:MPLS-LSR-STD-MIB?module=MPLS-LSR-STD-MIB&revision=2004-06-03
urn:ietf:params:xml:ns:yang:smiv2:MPLS-TC-MIB?module=MPLS-TC-MIB&revision=2001-01-04
urn:ietf:params:xml:ns:yang:smiv2:MPLS-TC-STD-MIB?module=MPLS-TC-STD-MIB&revision=2004-06-03
urn:ietf:params:xml:ns:yang:smiv2:MPLS-TE-STD-MIB?module=MPLS-TE-STD-MIB&revision=2004-06-03
urn:ietf:params:xml:ns:yang:smiv2:MPLS-VPN-MIB?module=MPLS-VPN-MIB&revision=2001-10-15
urn:ietf:params:xml:ns:yang:smiv2:NHRP-MIB?module=NHRP-MIB&revision=1999-08-26
urn:ietf:params:xml:ns:yang:smiv2:NOTIFICATION-LOG-MIB?module=NOTIFICATION-LOG-MIB&revision=2000-11-27
urn:ietf:params:xml:ns:yang:smiv2:OSPF-MIB?module=OSPF-MIB&revision=2006-11-10
urn:ietf:params:xml:ns:yang:smiv2:OSPF-TRAP-MIB?module=OSPF-TRAP-MIB&revision=2006-11-10
urn:ietf:params:xml:ns:yang:smiv2:P-BRIDGE-MIB?module=P-BRIDGE-MIB&revision=2006-01-09
urn:ietf:params:xml:ns:yang:smiv2:PIM-MIB?module=PIM-MIB&revision=2000-09-28
urn:ietf:params:xml:ns:yang:smiv2:PerfHist-TC-MIB?module=PerfHist-TC-MIB&revision=1998-11-07
urn:ietf:params:xml:ns:yang:smiv2:Q-BRIDGE-MIB?module=Q-BRIDGE-MIB&revision=2006-01-09
urn:ietf:params:xml:ns:yang:smiv2:RFC-1212?module=RFC-1212
urn:ietf:params:xml:ns:yang:smiv2:RFC-1215?module=RFC-1215
urn:ietf:params:xml:ns:yang:smiv2:RFC1155-SMI?module=RFC1155-SMI
urn:ietf:params:xml:ns:yang:smiv2:RFC1213-MIB?module=RFC1213-MIB
urn:ietf:params:xml:ns:yang:smiv2:RFC1315-MIB?module=RFC1315-MIB
urn:ietf:params:xml:ns:yang:smiv2:RMON-MIB?module=RMON-MIB&revision=2000-05-11
urn:ietf:params:xml:ns:yang:smiv2:RMON2-MIB?module=RMON2-MIB&revision=1996-05-27
urn:ietf:params:xml:ns:yang:smiv2:RSVP-MIB?module=RSVP-MIB&revision=1998-08-25
urn:ietf:params:xml:ns:yang:smiv2:SNMP-FRAMEWORK-MIB?module=SNMP-FRAMEWORK-MIB&revision=2002-10-14
urn:ietf:params:xml:ns:yang:smiv2:SNMP-PROXY-MIB?module=SNMP-PROXY-MIB&revision=2002-10-14
urn:ietf:params:xml:ns:yang:smiv2:SNMP-TARGET-MIB?module=SNMP-TARGET-MIB&revision=1998-08-04
urn:ietf:params:xml:ns:yang:smiv2:SNMPv2-MIB?module=SNMPv2-MIB&revision=2002-10-16
urn:ietf:params:xml:ns:yang:smiv2:SNMPv2-TC?module=SNMPv2-TC
urn:ietf:params:xml:ns:yang:smiv2:SONET-MIB?module=SONET-MIB&revision=2003-08-11
urn:ietf:params:xml:ns:yang:smiv2:TCP-MIB?module=TCP-MIB&revision=2005-02-18
urn:ietf:params:xml:ns:yang:smiv2:TOKEN-RING-RMON-MIB?module=TOKEN-RING-RMON-MIB
urn:ietf:params:xml:ns:yang:smiv2:TOKENRING-MIB?module=TOKENRING-MIB&revision=1994-10-23
urn:ietf:params:xml:ns:yang:smiv2:TUNNEL-MIB?module=TUNNEL-MIB&revision=2005-05-16
urn:ietf:params:xml:ns:yang:smiv2:UDP-MIB?module=UDP-MIB&revision=2005-05-20
urn:ietf:params:xml:ns:yang:smiv2:VPN-TC-STD-MIB?module=VPN-TC-STD-MIB&revision=2005-11-15
urn:ietf:params:xml:ns:netconf:base:1.0?module=ietf-netconf&revision=2011-06-01
urn:ietf:params:xml:ns:yang:ietf-netconf-with-defaults?module=ietf-netconf-with-defaults&revision=2011-06-01
urn:ietf:params:netconf:capability:notification:1.1
###Markdown
Ok, that's a bit messy, so let's tidy it up a bit and look, initially, at all the base netconf capabilities:
###Code
nc_caps = [c for c in m.server_capabilities if c.startswith('urn:ietf:params:netconf') and "smiv2" not in c]
for c in nc_caps:
print(c)
###Output
urn:ietf:params:netconf:base:1.0
urn:ietf:params:netconf:base:1.1
urn:ietf:params:netconf:capability:writable-running:1.0
urn:ietf:params:netconf:capability:xpath:1.0
urn:ietf:params:netconf:capability:validate:1.0
urn:ietf:params:netconf:capability:validate:1.1
urn:ietf:params:netconf:capability:rollback-on-error:1.0
urn:ietf:params:netconf:capability:notification:1.0
urn:ietf:params:netconf:capability:interleave:1.0
urn:ietf:params:netconf:capability:with-defaults:1.0?basic-mode=explicit&also-supported=report-all-tagged
urn:ietf:params:netconf:capability:yang-library:1.0?revision=2016-06-21&module-set-id=61de1a2313e60afe62df413a77ed9087
###Markdown
And now let's look at the capabilities that are related to model support:
###Code
import re
for c in m.server_capabilities:
model = re.search('module=([^&]*)&', c)
if model is not None:
print("{}".format(model.group(1)))
revision = re.search('revision=([0-9]{4}-[0-9]{2}-[0-9]{2})', c)
if revision is not None:
print(" revision = {}".format(revision.group(1)))
deviations = re.search('deviations=([a-zA-Z0-9\-,]+)($|&)',c)
if deviations is not None:
print(" deviations = {}".format(deviations.group(1)))
features = re.search('features=([a-zA-Z0-9\-,]+)($|&)',c)
if features is not None:
print(" features = {}".format(features.group(1)))
###Output
_____no_output_____
###Markdown
SchemaLet's take a look at playing with schema. First, we can try downloading them, picking one of the modules we got capabilities for.
###Code
# SCHEMA_TO_GET = 'openconfig-interfaces'
# SCHEMA_TO_GET = 'Cisco-IOS-XE-native'
# SCHEMA_TO_GET = 'Cisco-IOS-XE-features'
SCHEMA_TO_GET = 'Cisco-IOS-XE-interfaces-oper'
c = m.get_schema(SCHEMA_TO_GET)
print(c.data)
###Output
module Cisco-IOS-XE-interfaces-oper {
yang-version 1;
namespace "http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper";
prefix interfaces-ios-xe-oper;
import ietf-inet-types {
prefix inet;
}
import ietf-yang-types {
prefix yang;
}
import cisco-semver {
prefix cisco-semver;
}
organization
"Cisco Systems, Inc.";
contact
"Cisco Systems, Inc.
Customer Service
Postal: 170 W Tasman Drive
San Jose, CA 95134
Tel: +1 1800 553-NETS
E-mail: [email protected]";
description
"This module contains a collection of YANG definitions for
monitoring the interfaces in a Network Element.
Copyright (c) 2016-2019 by Cisco Systems, Inc.
All rights reserved.";
revision 2019-11-01 {
description
"- Added media type to interface
- Added more interface speed values";
reference "3.3.0";
cisco-semver:module-version "3.3.0";
}
revision 2019-05-01 {
description
"- Added ethernet error counters.
- Added semantic version";
reference "3.2.0";
cisco-semver:module-version "3.2.0";
}
revision 2018-10-29 {
description
"Cleaned up spelling errors in descriptions.";
reference "3.1.0";
cisco-semver:module-version "3.1.0";
}
revision 2018-06-29 {
description
"- Add switching interface speed values
- Add synchronous serial interface model";
reference "3.0.0";
cisco-semver:module-version "3.0.0";
}
revision 2018-02-01 {
description
"IPv6 addresses";
reference "2.0.0";
cisco-semver:module-version "2.0.0";
}
revision 2017-10-10 {
description
"Initial revision";
reference "1.0.0";
cisco-semver:module-version "1.0.0";
}
typedef qos-match-type {
type enumeration {
enum "qos-match-dscp" {
value 0;
}
enum "qos-match-src-ip" {
value 1;
}
enum "qos-match-dst-ip" {
value 2;
}
enum "qos-match-src-port" {
value 3;
}
enum "qos-match-dst-port" {
value 4;
}
enum "qos-match-proto" {
value 5;
}
}
description
"QOS match type";
}
typedef thresh-unit {
type enumeration {
enum "thresh-units-default" {
value 0;
}
enum "thresh-units-bytes" {
value 1;
}
enum "thresh-units-sec" {
value 2;
}
enum "thresh-units-packets" {
value 3;
}
enum "thresh-units-cells" {
value 4;
}
enum "thresh-units-percent" {
value 5;
}
}
description
"Units of threshold";
}
typedef qos-direction {
type enumeration {
enum "qos-inbound" {
value 0;
description
"Direction of traffic coming into the network entry";
}
enum "qos-outbound" {
value 1;
description
"Direction of traffic going out of the network entry";
}
}
description
"QoS direction indication";
}
typedef aggregation-type {
type enumeration {
enum "lag-off" {
value 0;
description
"LAG mode is off";
}
enum "lag-auto" {
value 1;
description
"LAG mode is auto";
}
enum "lag-active" {
value 2;
description
"LAG mode is active";
}
enum "lag-passive" {
value 3;
description
"LAG mode is passive";
}
}
description
"Type to define the lag-type, i.e., how the LAG is
defined and managed";
}
typedef intf-state {
type enumeration {
enum "if-state-unknown" {
value 0;
}
enum "if-state-up" {
value 1;
}
enum "if-state-down" {
value 2;
}
enum "if-state-test" {
value 3;
}
}
description
"The desired state of the interface.
This leaf has the same read semantics as ifAdminStatus.
Reference:
RFC 2863: The Interfaces Group MIB - ifAdminStatus";
}
typedef ether-duplex {
type enumeration {
enum "full-duplex" {
value 0;
}
enum "half-duplex" {
value 1;
}
enum "auto-duplex" {
value 2;
}
enum "unknown-duplex" {
value 3;
}
}
description
"The duplex setting of the interface";
}
typedef ether-speed {
type enumeration {
enum "speed-10mb" {
value 0;
}
enum "speed-100mb" {
value 1;
}
enum "speed-1gb" {
value 2;
}
enum "speed-10gb" {
value 3;
}
enum "speed-25gb" {
value 4;
}
enum "speed-40gb" {
value 5;
}
enum "speed-50gb" {
value 6;
}
enum "speed-100gb" {
value 7;
}
enum "speed-unknown" {
value 8;
}
enum "speed-auto" {
value 9;
}
enum "speed-2500mb" {
value 10;
description
"Ethernet Speed 2500 MBPS";
}
enum "speed-5gb" {
value 11;
description
"Ethernet Speed 5 GBPS";
}
enum "speed-400gb" {
value 12;
description
"Ethernet Speed 400 GBPS";
}
}
description
"The speed setting of the interface";
}
typedef oper-state {
type enumeration {
enum "if-oper-state-invalid" {
value 0;
}
enum "if-oper-state-ready" {
value 1;
}
enum "if-oper-state-no-pass" {
value 2;
}
enum "if-oper-state-test" {
value 3;
}
enum "if-oper-state-unknown" {
value 4;
}
enum "if-oper-state-dormant" {
value 5;
}
enum "if-oper-state-not-present" {
value 6;
}
enum "if-oper-state-lower-layer-down" {
value 7;
}
}
description
"The current operational state of the interface.
This leaf has the same semantics as ifOperStatus.
Reference:
RFC 2863: The Interfaces Group MIB - ifOperStatus";
}
typedef ietf-intf-type {
type enumeration {
enum "iana-iftype-other" {
value 1;
}
enum "iana-iftype-regular1822" {
value 2;
}
enum "iana-iftype-hdh1822" {
value 3;
}
enum "iana-iftype-ddnx25" {
value 4;
}
enum "iana-iftype-rfc877x25" {
value 5;
}
enum "iana-iftype-ethernet-csmacd" {
value 6;
}
enum "iana-iftype-iso88023-csmacd" {
value 7;
}
enum "iana-iftype-iso88024-tokenbus" {
value 8;
}
enum "iana-iftype-iso88025-tokenring" {
value 9;
}
enum "iana-iftype-iso88026-man" {
value 10;
}
enum "iana-iftype-starlan" {
value 11;
}
enum "iana-iftype-proteon10mbit" {
value 12;
}
enum "iana-iftype-proteon80mbit" {
value 13;
}
enum "iana-iftype-hyperchannel" {
value 14;
}
enum "iana-iftype-fddi" {
value 15;
}
enum "iana-iftype-lapb" {
value 16;
}
enum "iana-iftype-sdlc" {
value 17;
}
enum "iana-iftype-ds1" {
value 18;
}
enum "iana-iftype-e1" {
value 19;
}
enum "iana-iftype-basicisdn" {
value 20;
}
enum "iana-iftype-primaryisdn" {
value 21;
}
enum "iana-iftype-prop-p2p-serial" {
value 22;
}
enum "iana-iftype-ppp" {
value 23;
}
enum "iana-iftype-sw-loopback" {
value 24;
}
enum "iana-iftype-eon" {
value 25;
}
enum "iana-iftype-ethernet3mbit" {
value 26;
}
enum "iana-iftype-nsip" {
value 27;
}
enum "iana-iftype-slip" {
value 28;
}
enum "iana-iftype-ultra" {
value 29;
}
enum "iana-iftype-ds3" {
value 30;
}
enum "iana-iftype-sip" {
value 31;
}
enum "iana-iftype-framerelay" {
value 32;
}
enum "iana-iftype-rs232" {
value 33;
}
enum "iana-iftype-para" {
value 34;
}
enum "iana-iftype-arcnet" {
value 35;
}
enum "iana-iftype-arcnetplus" {
value 36;
}
enum "iana-iftype-atm" {
value 37;
}
enum "iana-iftype-miox25" {
value 38;
}
enum "iana-iftype-sonet" {
value 39;
}
enum "iana-iftype-x25ple" {
value 40;
}
enum "iana-iftype-iso88022-llc" {
value 41;
}
enum "iana-iftype-localtalk" {
value 42;
}
enum "iana-iftype-smdsdxi" {
value 43;
}
enum "iana-iftype-framerelay-service" {
value 44;
}
enum "iana-iftype-v35" {
value 45;
}
enum "iana-iftype-hssi" {
value 46;
}
enum "iana-iftype-hippi" {
value 47;
}
enum "iana-iftype-modem" {
value 48;
}
enum "iana-iftype-aal5" {
value 49;
}
enum "iana-iftype-sonetpath" {
value 50;
}
enum "iana-iftype-sonetvt" {
value 51;
}
enum "iana-iftype-smdsicip" {
value 52;
}
enum "iana-iftype-propvirtual" {
value 53;
}
enum "iana-iftype-propmultiplexor" {
value 54;
}
enum "iana-iftype-ieee80212" {
value 55;
}
enum "iana-iftype-fiberchannel" {
value 56;
}
enum "iana-iftype-hippi-interface" {
value 57;
}
enum "iana-iftype-framerelay-interconnect" {
value 58;
}
enum "iana-iftype-aflane8023" {
value 59;
}
enum "iana-iftype-aflane8025" {
value 60;
}
enum "iana-iftype-cctemul" {
value 61;
}
enum "iana-iftype-fastether" {
value 62;
}
enum "iana-iftype-isdn" {
value 63;
}
enum "iana-iftype-v11" {
value 64;
}
enum "iana-iftype-v36" {
value 65;
}
enum "iana-iftype-g703at64k" {
value 66;
}
enum "iana-iftype-g703at2mb" {
value 67;
}
enum "iana-iftype-qllc" {
value 68;
}
enum "iana-iftype-fastetherfx" {
value 69;
}
enum "iana-iftype-channel" {
value 70;
}
enum "iana-iftype-ieee80211" {
value 71;
}
enum "iana-iftype-ibm370parchan" {
value 72;
}
enum "iana-iftype-escon" {
value 73;
}
enum "iana-iftype-dlsw" {
value 74;
}
enum "iana-iftype-isdns" {
value 75;
}
enum "iana-iftype-isdnu" {
value 76;
}
enum "iana-iftype-lapd" {
value 77;
}
enum "iana-iftype-ipswitch" {
value 78;
}
enum "iana-iftype-rsrb" {
value 79;
}
enum "iana-iftype-atmlogical" {
value 80;
}
enum "iana-iftype-ds0" {
value 81;
}
enum "iana-iftype-ds0bundle" {
value 82;
}
enum "iana-iftype-bsc" {
value 83;
}
enum "iana-iftype-async" {
value 84;
}
enum "iana-iftype-cnr" {
value 85;
}
enum "iana-iftype-iso88025-dtr" {
value 86;
}
enum "iana-iftype-eplrs" {
value 87;
}
enum "iana-iftype-arap" {
value 88;
}
enum "iana-iftype-propcnls" {
value 89;
}
enum "iana-iftype-hostpad" {
value 90;
}
enum "iana-iftype-termpad" {
value 91;
}
enum "iana-iftype-framerelay-mpi" {
value 92;
}
enum "iana-iftype-x213" {
value 93;
}
enum "iana-iftype-adsl" {
value 94;
}
enum "iana-iftype-radsl" {
value 95;
}
enum "iana-iftype-sdsl" {
value 96;
}
enum "iana-iftype-vdsl" {
value 97;
}
enum "iana-iftype-iso88025-crfpint" {
value 98;
}
enum "iana-iftype-myrinet" {
value 99;
}
enum "iana-iftype-voiceem" {
value 100;
}
enum "iana-iftype-voicefxo" {
value 101;
}
enum "iana-iftype-voicefxs" {
value 102;
}
enum "iana-iftype-voiceencap" {
value 103;
}
enum "iana-iftype-voip" {
value 104;
}
enum "iana-iftype-atmdxi" {
value 105;
}
enum "iana-iftype-atmfuni" {
value 106;
}
enum "iana-iftype-atmima" {
value 107;
}
enum "iana-iftype-ppp-multilinkbundle" {
value 108;
}
enum "iana-iftype-ipovercdlc" {
value 109;
}
enum "iana-iftype-ipoverclaw" {
value 110;
}
enum "iana-iftype-stack2stack" {
value 111;
}
enum "iana-iftype-virtualipaddress" {
value 112;
}
enum "iana-iftype-mpc" {
value 113;
}
enum "iana-iftype-ipoveratm" {
value 114;
}
enum "iana-iftype-iso88025-fiber" {
value 115;
}
enum "iana-iftype-tdlc" {
value 116;
}
enum "iana-iftype-gige" {
value 117;
}
enum "iana-iftype-hdlc" {
value 118;
}
enum "iana-iftype-lapf" {
value 119;
}
enum "iana-iftype-v37" {
value 120;
}
enum "iana-iftype-x25mlp" {
value 121;
}
enum "iana-iftype-x25huntgroup" {
value 122;
}
enum "iana-iftype-transphdlc" {
value 123;
}
enum "iana-iftype-interleave" {
value 124;
}
enum "iana-iftype-fast" {
value 125;
}
enum "iana-iftype-ip" {
value 126;
}
enum "iana-iftype-docs-cable-maclayer" {
value 127;
}
enum "iana-iftype-docs-cable-downstream" {
value 128;
}
enum "iana-iftype-docs-cable-upstream" {
value 129;
}
enum "iana-iftype-a12mppswitch" {
value 130;
}
enum "iana-iftype-tunnel" {
value 131;
}
enum "iana-iftype-coffee" {
value 132;
}
enum "iana-iftype-ces" {
value 133;
}
enum "iana-iftype-atmsubinterface" {
value 134;
}
enum "iana-iftype-l2vlan" {
value 135;
}
enum "iana-iftype-l3ipvlan" {
value 136;
}
enum "iana-iftype-l3ipxvlan" {
value 137;
}
enum "iana-iftype-digital-powerline" {
value 138;
}
enum "iana-iftype-media-mailoverip" {
value 139;
}
enum "iana-iftype-dtm" {
value 140;
}
enum "iana-iftype-dcn" {
value 141;
}
enum "iana-iftype-ipforward" {
value 142;
}
enum "iana-iftype-msdsl" {
value 143;
}
enum "iana-iftype-ieee1394" {
value 144;
}
enum "iana-iftype-gsn" {
value 145;
}
enum "iana-iftype-dvbrcc-maclayer" {
value 146;
}
enum "iana-iftype-dvbrcc-downstream" {
value 147;
}
enum "iana-iftype-dvbrcc-upstream" {
value 148;
}
enum "iana-iftype-atmvirtual" {
value 149;
}
enum "iana-iftype-mplstunnel" {
value 150;
}
enum "iana-iftype-srp" {
value 151;
}
enum "iana-iftype-voiceoveratm" {
value 152;
}
enum "iana-iftype-voiceoverframerelay" {
value 153;
}
enum "iana-iftype-idsl" {
value 154;
}
enum "iana-iftype-compositelink" {
value 155;
}
enum "iana-iftype-ss7siglink" {
value 156;
}
enum "iana-iftype-propwireless-p2p" {
value 157;
}
enum "iana-iftype-frforward" {
value 158;
}
enum "iana-iftype-rfc1483" {
value 159;
}
enum "iana-iftype-usb" {
value 160;
}
enum "iana-iftype-ieee8023-adlag" {
value 161;
}
enum "iana-iftype-bgppolicy-accounting" {
value 162;
}
enum "iana-iftype-frf16mfrbundle" {
value 163;
}
enum "iana-iftype-h323gatekeeper" {
value 164;
}
enum "iana-iftype-h323proxy" {
value 165;
}
enum "iana-iftype-mpls" {
value 166;
}
enum "iana-iftype-mfsiglink" {
value 167;
}
enum "iana-iftype-hdsl2" {
value 168;
}
enum "iana-iftype-shdsl" {
value 169;
}
enum "iana-iftype-ds1fdl" {
value 170;
}
enum "iana-iftype-pos" {
value 171;
}
enum "iana-iftype-dvbasiin" {
value 172;
}
enum "iana-iftype-dvbasiout" {
value 173;
}
enum "iana-iftype-plc" {
value 174;
}
enum "iana-iftype-nfas" {
value 175;
}
enum "iana-iftype-tr008" {
value 176;
}
enum "iana-iftype-gr303rdt" {
value 177;
}
enum "iana-iftype-gr303idt" {
value 178;
}
enum "iana-iftype-isup" {
value 179;
}
enum "iana-iftype-prop-docs-wireless-maclayer" {
value 180;
}
enum "iana-iftype-prop-docs-wireless-downstream" {
value 181;
}
enum "iana-iftype-prop-docs-wireless-upstream" {
value 182;
}
enum "iana-iftype-hiperlan2" {
value 183;
}
enum "iana-iftype-prop-bwap2mp" {
value 184;
}
enum "iana-iftype-sonetoverheadchannel" {
value 185;
}
enum "iana-iftype-digital-wrapperoverheadchannel" {
value 186;
}
enum "iana-iftype-aal2" {
value 187;
}
enum "iana-iftype-radiomac" {
value 188;
}
enum "iana-iftype-atmradio" {
value 189;
}
enum "iana-iftype-imt" {
value 190;
}
enum "iana-iftype-mvl" {
value 191;
}
enum "iana-iftype-reachhdsl" {
value 192;
}
enum "iana-iftype-frdlciendpt" {
value 193;
}
enum "iana-iftype-atmvciendpt" {
value 194;
}
enum "iana-iftype-opticalchannel" {
value 195;
}
enum "iana-iftype-opticaltransport" {
value 196;
}
enum "iana-iftype-propatm" {
value 197;
}
enum "iana-iftype-voiceovercable" {
value 198;
}
enum "iana-iftype-infiniband" {
value 199;
}
enum "iana-iftype-telink" {
value 200;
}
enum "iana-iftype-q2931" {
value 201;
}
enum "iana-iftype-virtualatg" {
value 202;
}
enum "iana-iftype-siptg" {
value 203;
}
enum "iana-iftype-sipsig" {
value 204;
}
enum "iana-iftype-docs-cable-upstreamchannel" {
value 205;
}
enum "iana-iftype-econet" {
value 206;
}
enum "iana-iftype-pon155" {
value 207;
}
enum "iana-iftype-pon622" {
value 208;
}
enum "iana-iftype-bridge-if" {
value 209;
}
enum "iana-iftype-linegroup" {
value 210;
}
enum "iana-iftype-voiceemfgd" {
value 211;
}
enum "iana-iftype-voiceefgdeana" {
value 212;
}
enum "iana-iftype-voicedid" {
value 213;
}
enum "iana-iftype-mpegtransport" {
value 214;
}
enum "iana-iftype-sixtofour" {
value 215;
}
enum "iana-iftype-gtp" {
value 216;
}
enum "iana-iftype-pdnetherloop1" {
value 217;
}
enum "iana-iftype-pdnetherloop2" {
value 218;
}
enum "iana-iftype-opticalchannel-group" {
value 219;
}
enum "iana-iftype-homepna" {
value 220;
}
enum "iana-iftype-gfp" {
value 221;
}
enum "iana-iftype-ciscoislvlan" {
value 222;
}
enum "iana-iftype-actelismetaloop" {
value 223;
}
enum "iana-iftype-fciplink" {
value 224;
}
enum "iana-iftype-rpr" {
value 225;
}
enum "iana-iftype-qam" {
value 226;
}
enum "iana-iftype-lmp" {
value 227;
}
enum "iana-iftype-cblvectastar" {
value 228;
}
enum "iana-iftype-docs-cable-mcmts-downtream" {
value 229;
}
enum "iana-iftype-adsl2" {
value 230;
}
enum "iana-iftype-macseccontrolledif" {
value 231;
}
enum "iana-iftype-macsecuncontrolledif" {
value 232;
}
enum "iana-iftype-aviciopticalether" {
value 233;
}
enum "iana-iftype-atmbond" {
value 234;
}
enum "iana-iftype-voicefgdos" {
value 235;
}
enum "iana-iftype-mocaversion1" {
value 236;
}
enum "iana-iftype-ieee80216-wman" {
value 237;
}
enum "iana-iftype-adsl2plus" {
value 238;
}
enum "iana-iftype-dvbrcsmaclayer" {
value 239;
}
enum "iana-iftype-dvbtdm" {
value 240;
}
enum "iana-iftype-dvbrcstdma" {
value 241;
}
enum "iana-iftype-x86laps" {
value 242;
}
enum "iana-iftype-wwanpp" {
value 243;
}
enum "iana-iftype-wwanpp2" {
value 244;
}
enum "iana-iftype-voiceebs" {
value 245;
}
enum "iana-iftype-ifpwtype" {
value 246;
}
enum "iana-iftype-ilan" {
value 247;
}
enum "iana-iftype-pip" {
value 248;
}
enum "iana-iftype-aluelp" {
value 249;
}
enum "iana-iftype-gpon" {
value 250;
}
enum "iana-iftype-vdsl2" {
value 251;
}
enum "iana-iftype-capwapdot11-profile" {
value 252;
}
enum "iana-iftype-capwapdot11-bss" {
value 253;
}
enum "iana-iftype-capwapwtp-virtualradio" {
value 254;
}
enum "iana-iftype-bits" {
value 255;
}
enum "iana-iftype-docs-cable-upstreamrfport" {
value 256;
}
enum "iana-iftype-cable-downstreamrfport" {
value 257;
}
enum "iana-iftype-vmware-virtualnic" {
value 258;
}
enum "iana-iftype-ieee802154" {
value 259;
}
enum "iana-iftype-otnodu" {
value 260;
}
enum "iana-iftype-otnotu" {
value 261;
}
enum "iana-iftype-ifvfitype" {
value 262;
}
enum "iana-iftype-g9981" {
value 263;
}
enum "iana-iftype-g9982" {
value 264;
}
enum "iana-iftype-g9983" {
value 265;
}
enum "iana-iftype-aluepon" {
value 266;
}
enum "iana-iftype-aluepon-onu" {
value 267;
}
enum "iana-iftype-aluepon-physicaluni" {
value 268;
}
enum "iana-iftype-aluepon-logicalink" {
value 269;
}
enum "iana-iftype-alugpon-onu" {
value 270;
}
enum "iana-iftype-alugpon-physicaluni" {
value 271;
}
enum "iana-iftype-vmwarenicteam" {
value 272;
}
enum "iana-iftype-docs-ofdm-downstream" {
value 277;
}
enum "iana-iftype-docs-ofdma-upstream" {
value 278;
}
enum "iana-iftype-gfast" {
value 279;
}
enum "iana-iftype-sdci" {
value 280;
}
enum "iana-iftype-xbox-wireless" {
value 281;
}
enum "iana-iftype-fastdsl" {
value 282;
}
}
description
"IANAifType
This data type is used as the syntax of the ifType
object in the (updated) definition of MIB-II's ifTable";
}
typedef media-type-class {
type enumeration {
enum "ether-media-type-none" {
value 0;
}
enum "ether-media-type-aui" {
value 1;
}
enum "ether-media-type-10baset" {
value 2;
}
enum "ether-media-type-nothing-here" {
value 3;
}
enum "ether-media-type-rj45" {
value 4;
}
enum "ether-media-type-mii" {
value 5;
}
enum "ether-media-type-auto-select" {
value 6;
}
enum "ether-media-type-epif-utp" {
value 7;
}
enum "ether-media-type-epif-fx" {
value 8;
}
enum "ether-media-type-gpif-lx" {
value 9;
}
enum "ether-media-type-gpif-sx" {
value 10;
}
enum "ether-media-type-parallel" {
value 11;
}
enum "ether-media-type-serial10km" {
value 12;
}
enum "ether-media-type-vsr300m" {
value 13;
}
enum "ether-media-type-metrowdm50km" {
value 14;
}
enum "ether-media-type-not-connected" {
value 15;
}
enum "ether-media-type-port-missing" {
value 16;
}
enum "ether-media-type-gbic-missing" {
value 17;
}
enum "ether-media-type-feip-mmf-st" {
value 18;
}
enum "ether-media-type-feip-mmf-sc" {
value 19;
}
enum "ether-media-type-1000basecx-gigstack" {
value 20;
}
enum "ether-media-type-serial40km" {
value 21;
}
enum "ether-media-type-serial50km" {
value 22;
}
enum "ether-media-type-unapproved" {
value 23;
}
enum "ether-media-type-unsupported" {
value 24;
}
enum "ether-media-type-bad-eeprom" {
value 25;
}
enum "ether-media-type-gbic" {
value 26;
}
enum "ether-media-type-sfp" {
value 27;
}
enum "ether-media-type-xenpak" {
value 28;
}
enum "ether-media-type-100base-fx" {
value 29;
}
enum "ether-media-type-100base-tx" {
value 30;
}
enum "ether-media-type-rj21" {
value 31;
}
enum "ether-media-type-100mtrj-fx" {
value 32;
}
enum "ether-media-type-100mtrj-lx" {
value 33;
}
enum "ether-media-type-internal" {
value 34;
}
enum "ether-media-type-1000baset" {
value 35;
}
enum "ether-media-type-gpif-zx" {
value 36;
}
enum "ether-media-type-1000base-cwdm-1470" {
value 37;
}
enum "ether-media-type-1000base-cwdm-1490" {
value 38;
}
enum "ether-media-type-1000base-cwdm-1510" {
value 39;
}
enum "ether-media-type-1000base-cwdm-1530" {
value 40;
}
enum "ether-media-type-1000base-cwdm-1550" {
value 41;
}
enum "ether-media-type-1000base-cwdm-1570" {
value 42;
}
enum "ether-media-type-1000base-cwdm-1590" {
value 43;
}
enum "ether-media-type-1000base-cwdm-1610" {
value 44;
}
enum "ether-media-type-1000base-dwdm-6061" {
value 45;
}
enum "ether-media-type-1000base-dwdm-5979" {
value 46;
}
enum "ether-media-type-1000base-dwdm-5898" {
value 47;
}
enum "ether-media-type-1000base-dwdm-5817" {
value 48;
}
enum "ether-media-type-1000base-dwdm-5655" {
value 49;
}
enum "ether-media-type-1000base-dwdm-5575" {
value 50;
}
enum "ether-media-type-1000base-dwdm-5494" {
value 51;
}
enum "ether-media-type-1000base-dwdm-5413" {
value 52;
}
enum "ether-media-type-1000base-dwdm-5252" {
value 53;
}
enum "ether-media-type-1000base-dwdm-5172" {
value 54;
}
enum "ether-media-type-1000base-dwdm-5092" {
value 55;
}
enum "ether-media-type-1000base-dwdm-5012" {
value 56;
}
enum "ether-media-type-1000base-dwdm-4851" {
value 57;
}
enum "ether-media-type-1000base-dwdm-4772" {
value 58;
}
enum "ether-media-type-1000base-dwdm-4692" {
value 59;
}
enum "ether-media-type-1000base-dwdm-4612" {
value 60;
}
enum "ether-media-type-1000base-dwdm-4453" {
value 61;
}
enum "ether-media-type-1000base-dwdm-4373" {
value 62;
}
enum "ether-media-type-1000base-dwdm-4294" {
value 63;
}
enum "ether-media-type-1000base-dwdm-4214" {
value 64;
}
enum "ether-media-type-1000base-dwdm-4056" {
value 65;
}
enum "ether-media-type-1000base-dwdm-3977" {
value 66;
}
enum "ether-media-type-1000base-dwdm-3898" {
value 67;
}
enum "ether-media-type-1000base-dwdm-3819" {
value 68;
}
enum "ether-media-type-1000base-dwdm-3661" {
value 69;
}
enum "ether-media-type-1000base-dwdm-3582" {
value 70;
}
enum "ether-media-type-1000base-dwdm-3504" {
value 71;
}
enum "ether-media-type-1000base-dwdm-3425" {
value 72;
}
enum "ether-media-type-1000base-dwdm-3268" {
value 73;
}
enum "ether-media-type-1000base-dwdm-3190" {
value 74;
}
enum "ether-media-type-1000base-dwdm-3112" {
value 75;
}
enum "ether-media-type-1000base-dwdm-3033" {
value 76;
}
enum "ether-media-type-1000base-dwdm-6141" {
value 77;
}
enum "ether-media-type-1000base-dwdm-5736" {
value 78;
}
enum "ether-media-type-1000base-dwdm-5332" {
value 79;
}
enum "ether-media-type-1000base-dwdm-4931" {
value 80;
}
enum "ether-media-type-1000base-dwdm-4532" {
value 81;
}
enum "ether-media-type-1000base-dwdm-4134" {
value 82;
}
enum "ether-media-type-1000base-dwdm-3739" {
value 83;
}
enum "ether-media-type-1000base-dwdm-3346" {
value 84;
}
enum "ether-media-type-unknown" {
value 85;
}
enum "ether-media-type-1000base-bx10u" {
value 86;
}
enum "ether-media-type-1000base-bx10d" {
value 87;
}
enum "ether-media-type-1000base-tdwdm" {
value 88;
}
enum "ether-media-type-10gbase-sr" {
value 89;
}
enum "ether-media-type-10gbase-sr-sw" {
value 90;
}
enum "ether-media-type-sfp-plus" {
value 91;
}
enum "ether-media-type-gpif-ex" {
value 92;
}
enum "ether-media-type-100base-bx10u" {
value 93;
}
enum "ether-media-type-100base-bx10d" {
value 94;
}
enum "ether-media-type-10gbase-dwdm-3033" {
value 95;
}
enum "ether-media-type-10gbase-dwdm-3112" {
value 96;
}
enum "ether-media-type-10gbase-dwdm-3190" {
value 97;
}
enum "ether-media-type-10gbase-dwdm-3268" {
value 98;
}
enum "ether-media-type-10gbase-dwdm-3425" {
value 99;
}
enum "ether-media-type-10gbase-dwdm-3504" {
value 100;
}
enum "ether-media-type-10gbase-dwdm-3582" {
value 101;
}
enum "ether-media-type-10gbase-dwdm-3661" {
value 102;
}
enum "ether-media-type-10gbase-dwdm-3819" {
value 103;
}
enum "ether-media-type-10gbase-dwdm-3898" {
value 104;
}
enum "ether-media-type-10gbase-dwdm-3977" {
value 105;
}
enum "ether-media-type-10gbase-dwdm-4056" {
value 106;
}
enum "ether-media-type-10gbase-dwdm-4214" {
value 107;
}
enum "ether-media-type-10gbase-dwdm-4294" {
value 108;
}
enum "ether-media-type-10gbase-dwdm-4373" {
value 109;
}
enum "ether-media-type-10gbase-dwdm-4453" {
value 110;
}
enum "ether-media-type-10gbase-dwdm-4612" {
value 111;
}
enum "ether-media-type-10gbase-dwdm-4692" {
value 112;
}
enum "ether-media-type-10gbase-dwdm-4772" {
value 113;
}
enum "ether-media-type-10gbase-dwdm-4851" {
value 114;
}
enum "ether-media-type-10gbase-dwdm-5012" {
value 115;
}
enum "ether-media-type-10gbase-dwdm-5092" {
value 116;
}
enum "ether-media-type-10gbase-dwdm-5172" {
value 117;
}
enum "ether-media-type-10gbase-dwdm-5252" {
value 118;
}
enum "ether-media-type-10gbase-dwdm-5413" {
value 119;
}
enum "ether-media-type-10gbase-dwdm-5494" {
value 120;
}
enum "ether-media-type-10gbase-dwdm-5575" {
value 121;
}
enum "ether-media-type-10gbase-dwdm-5655" {
value 122;
}
enum "ether-media-type-10gbase-dwdm-5817" {
value 123;
}
enum "ether-media-type-10gbase-dwdm-5898" {
value 124;
}
enum "ether-media-type-10gbase-dwdm-5979" {
value 125;
}
enum "ether-media-type-10gbase-dwdm-6061" {
value 126;
}
enum "ether-media-type-10gbase-tdwdm" {
value 127;
}
enum "ether-media-type-100base-zx" {
value 128;
}
enum "ether-media-type-100base-ex" {
value 129;
}
enum "ether-media-type-sfp-plus-10ge-sr" {
value 130;
}
enum "ether-media-type-sfp-plus-10ge-lr" {
value 131;
}
enum "ether-media-type-sfp-plus-10ge-lrm" {
value 132;
}
enum "ether-media-type-sfp-plus-10ge-er" {
value 133;
}
enum "ether-media-type-10gbase-cu1m" {
value 134;
}
enum "ether-media-type-10gbase-cu3m" {
value 135;
}
enum "ether-media-type-10gbase-cu5m" {
value 136;
}
enum "ether-media-type-10gbase-cu7m" {
value 137;
}
enum "ether-media-type-10gbase-cwdm-1470" {
value 138;
}
enum "ether-media-type-10gbase-cwdm-1490" {
value 139;
}
enum "ether-media-type-10gbase-cwdm-1510" {
value 140;
}
enum "ether-media-type-10gbase-cwdm-1530" {
value 141;
}
enum "ether-media-type-10gbase-cwdm-1550" {
value 142;
}
enum "ether-media-type-10gbase-cwdm-1570" {
value 143;
}
enum "ether-media-type-10gbase-cwdm-1590" {
value 144;
}
enum "ether-media-type-10gbase-cwdm-1610" {
value 145;
}
enum "ether-media-type-100base-lx" {
value 146;
}
enum "ether-media-type-100lc-bx10-d" {
value 147;
}
enum "ether-media-type-100lc-bx10-u" {
value 148;
}
enum "ether-media-type-100base-bx10-d" {
value 149;
}
enum "ether-media-type-100base-bx10-u" {
value 150;
}
enum "ether-media-type-x2" {
value 151;
}
enum "ether-media-type-x2-missing" {
value 152;
}
enum "ether-media-type-xg-draught" {
value 153;
}
enum "ether-media-type-1g-10gbaset" {
value 154;
}
enum "ether-media-type-xg-arcadia" {
value 155;
}
enum "ether-media-type-sfp-plus-10ge-zr" {
value 156;
}
enum "ether-media-type-sfpp-10g-dwdm-61-41" {
value 157;
}
enum "ether-media-type-sfpp-10g-dwdm-60-61" {
value 158;
}
enum "ether-media-type-sfpp-10g-dwdm-59-79" {
value 159;
}
enum "ether-media-type-sfpp-10g-dwdm-58-98" {
value 160;
}
enum "ether-media-type-sfpp-10g-dwdm-58-17" {
value 161;
}
enum "ether-media-type-sfpp-10g-dwdm-57-36" {
value 162;
}
enum "ether-media-type-sfpp-10g-dwdm-56-55" {
value 163;
}
enum "ether-media-type-sfpp-10g-dwdm-55-75" {
value 164;
}
enum "ether-media-type-sfpp-10g-dwdm-54-94" {
value 165;
}
enum "ether-media-type-sfpp-10g-dwdm-54-13" {
value 166;
}
enum "ether-media-type-sfpp-10g-dwdm-53-33" {
value 167;
}
enum "ether-media-type-sfpp-10g-dwdm-52-52" {
value 168;
}
enum "ether-media-type-sfpp-10g-dwdm-51-72" {
value 169;
}
enum "ether-media-type-sfpp-10g-dwdm-50-92" {
value 170;
}
enum "ether-media-type-sfpp-10g-dwdm-50-12" {
value 171;
}
enum "ether-media-type-sfpp-10g-dwdm-49-32" {
value 172;
}
enum "ether-media-type-sfpp-10g-dwdm-48-51" {
value 173;
}
enum "ether-media-type-sfpp-10g-dwdm-47-72" {
value 174;
}
enum "ether-media-type-sfpp-10g-dwdm-46-92" {
value 175;
}
enum "ether-media-type-sfpp-10g-dwdm-46-12" {
value 176;
}
enum "ether-media-type-sfpp-10g-dwdm-45-32" {
value 177;
}
enum "ether-media-type-sfpp-10g-dwdm-44-53" {
value 178;
}
enum "ether-media-type-sfpp-10g-dwdm-43-73" {
value 179;
}
enum "ether-media-type-sfpp-10g-dwdm-42-94" {
value 180;
}
enum "ether-media-type-sfpp-10g-dwdm-42-14" {
value 181;
}
enum "ether-media-type-sfpp-10g-dwdm-41-35" {
value 182;
}
enum "ether-media-type-sfpp-10g-dwdm-40-56" {
value 183;
}
enum "ether-media-type-sfpp-10g-dwdm-39-77" {
value 184;
}
enum "ether-media-type-sfpp-10g-dwdm-38-98" {
value 185;
}
enum "ether-media-type-sfpp-10g-dwdm-38-19" {
value 186;
}
enum "ether-media-type-sfpp-10g-dwdm-37-40" {
value 187;
}
enum "ether-media-type-sfpp-10g-dwdm-36-61" {
value 188;
}
enum "ether-media-type-sfpp-10g-dwdm-35-82" {
value 189;
}
enum "ether-media-type-sfpp-10g-dwdm-35-04" {
value 190;
}
enum "ether-media-type-sfpp-10g-dwdm-34-25" {
value 191;
}
enum "ether-media-type-sfpp-10g-dwdm-33-47" {
value 192;
}
enum "ether-media-type-sfpp-10g-dwdm-32-68" {
value 193;
}
enum "ether-media-type-sfpp-10g-dwdm-31-90" {
value 194;
}
enum "ether-media-type-sfpp-10g-dwdm-31-12" {
value 195;
}
enum "ether-media-type-sfpp-10g-dwdm-30-33" {
value 196;
}
enum "ether-media-type-1000base-bx40da" {
value 197;
}
enum "ether-media-type-1000base-bx80u" {
value 198;
}
enum "ether-media-type-1000base-bx80d" {
value 199;
}
enum "ether-media-type-1000base-bx40u" {
value 200;
}
enum "ether-media-type-1000base-bx40d" {
value 201;
}
enum "ether-media-type-sfp-plus-10ge-bxd" {
value 202;
}
enum "ether-media-type-sfp-plus-10ge-bxu" {
value 203;
}
enum "ether-media-type-sfp-plus-10ge-bx40d" {
value 204;
}
enum "ether-media-type-sfp-plus-10ge-bx40u" {
value 205;
}
enum "ether-media-type-sfpp-10g-cwdm-14-70" {
value 206;
}
enum "ether-media-type-sfpp-10g-cwdm-14-90" {
value 207;
}
enum "ether-media-type-sfpp-10g-cwdm-15-10" {
value 208;
}
enum "ether-media-type-sfpp-10g-cwdm-15-30" {
value 209;
}
enum "ether-media-type-sfpp-10g-cwdm-15-50" {
value 210;
}
enum "ether-media-type-sfpp-10g-cwdm-15-70" {
value 211;
}
enum "ether-media-type-sfpp-10g-cwdm-15-90" {
value 212;
}
enum "ether-media-type-sfpp-10g-cwdm-16-10" {
value 213;
}
enum "ether-media-type-sfpp-10g-acu7m" {
value 214;
}
enum "ether-media-type-sfpp-10g-acu10m" {
value 215;
}
enum "ether-media-type-cpak-100ge-sr10" {
value 216;
}
enum "ether-media-type-cpak-100ge-lr4" {
value 217;
}
enum "ether-media-type-cpak-100ge-sr4" {
value 218;
}
enum "ether-media-type-cfp-40g" {
value 219;
}
enum "ether-media-type-qsfp-40ge-lr" {
value 220;
}
enum "ether-media-type-qsfp-40ge-sr" {
value 221;
}
enum "ether-media-type-1000base-dr-lx" {
value 222;
}
enum "ether-media-type-sfpp-10g-tdwdm" {
value 223;
}
enum "ether-media-type-qsfp-40ge-er4" {
value 224;
}
enum "ether-media-type-sfpp-dwdm-10gep-61-83" {
value 225;
}
enum "ether-media-type-sfpp-dwdm-10gep-61-41" {
value 226;
}
enum "ether-media-type-sfpp-dwdm-10gep-61-01" {
value 227;
}
enum "ether-media-type-sfpp-dwdm-10gep-60-61" {
value 228;
}
enum "ether-media-type-sfpp-dwdm-10gep-60-20" {
value 229;
}
enum "ether-media-type-sfpp-dwdm-10gep-59-79" {
value 230;
}
enum "ether-media-type-sfpp-dwdm-10gep-59-39" {
value 231;
}
enum "ether-media-type-sfpp-dwdm-10gep-58-98" {
value 232;
}
enum "ether-media-type-sfpp-dwdm-10gep-58-58" {
value 233;
}
enum "ether-media-type-sfpp-dwdm-10gep-58-17" {
value 234;
}
enum "ether-media-type-sfpp-dwdm-10gep-57-77" {
value 235;
}
enum "ether-media-type-sfpp-dwdm-10gep-57-36" {
value 236;
}
enum "ether-media-type-sfpp-dwdm-10gep-56-96" {
value 237;
}
enum "ether-media-type-sfpp-dwdm-10gep-56-55" {
value 238;
}
enum "ether-media-type-sfpp-dwdm-10gep-56-15" {
value 239;
}
enum "ether-media-type-sfpp-dwdm-10gep-55-75" {
value 240;
}
enum "ether-media-type-sfpp-dwdm-10gep-55-34" {
value 241;
}
enum "ether-media-type-sfpp-dwdm-10gep-54-94" {
value 242;
}
enum "ether-media-type-sfpp-dwdm-10gep-54-54" {
value 243;
}
enum "ether-media-type-sfpp-dwdm-10gep-54-13" {
value 244;
}
enum "ether-media-type-sfpp-dwdm-10gep-53-73" {
value 245;
}
enum "ether-media-type-sfpp-dwdm-10gep-53-33" {
value 246;
}
enum "ether-media-type-sfpp-dwdm-10gep-52-93" {
value 247;
}
enum "ether-media-type-sfpp-dwdm-10gep-52-52" {
value 248;
}
enum "ether-media-type-sfpp-dwdm-10gep-52-12" {
value 249;
}
enum "ether-media-type-sfpp-dwdm-10gep-51-72" {
value 250;
}
enum "ether-media-type-sfpp-dwdm-10gep-51-32" {
value 251;
}
enum "ether-media-type-sfpp-dwdm-10gep-50-92" {
value 252;
}
enum "ether-media-type-sfpp-dwdm-10gep-50-52" {
value 253;
}
enum "ether-media-type-sfpp-dwdm-10gep-50-12" {
value 254;
}
enum "ether-media-type-sfpp-dwdm-10gep-49-72" {
value 255;
}
enum "ether-media-type-sfpp-dwdm-10gep-49-32" {
value 256;
}
enum "ether-media-type-sfpp-dwdm-10gep-48-91" {
value 257;
}
enum "ether-media-type-sfpp-dwdm-10gep-48-51" {
value 258;
}
enum "ether-media-type-sfpp-dwdm-10gep-48-11" {
value 259;
}
enum "ether-media-type-sfpp-dwdm-10gep-47-72" {
value 260;
}
enum "ether-media-type-sfpp-dwdm-10gep-47-32" {
value 261;
}
enum "ether-media-type-sfpp-dwdm-10gep-46-92" {
value 262;
}
enum "ether-media-type-sfpp-dwdm-10gep-46-52" {
value 263;
}
enum "ether-media-type-sfpp-dwdm-10gep-46-12" {
value 264;
}
enum "ether-media-type-sfpp-dwdm-10gep-45-72" {
value 265;
}
enum "ether-media-type-sfpp-dwdm-10gep-45-32" {
value 266;
}
enum "ether-media-type-sfpp-dwdm-10gep-44-92" {
value 267;
}
enum "ether-media-type-sfpp-dwdm-10gep-44-53" {
value 268;
}
enum "ether-media-type-sfpp-dwdm-10gep-44-13" {
value 269;
}
enum "ether-media-type-sfpp-dwdm-10gep-43-73" {
value 270;
}
enum "ether-media-type-sfpp-dwdm-10gep-43-33" {
value 271;
}
enum "ether-media-type-sfpp-dwdm-10gep-42-94" {
value 272;
}
enum "ether-media-type-sfpp-dwdm-10gep-42-54" {
value 273;
}
enum "ether-media-type-sfpp-dwdm-10gep-42-14" {
value 274;
}
enum "ether-media-type-sfpp-dwdm-10gep-41-75" {
value 275;
}
enum "ether-media-type-sfpp-dwdm-10gep-41-35" {
value 276;
}
enum "ether-media-type-sfpp-dwdm-10gep-40-95" {
value 277;
}
enum "ether-media-type-sfpp-dwdm-10gep-40-56" {
value 278;
}
enum "ether-media-type-sfpp-dwdm-10gep-40-16" {
value 279;
}
enum "ether-media-type-sfpp-dwdm-10gep-39-77" {
value 280;
}
enum "ether-media-type-sfpp-dwdm-10gep-39-37" {
value 281;
}
enum "ether-media-type-sfpp-dwdm-10gep-38-98" {
value 282;
}
enum "ether-media-type-sfpp-dwdm-10gep-38-58" {
value 283;
}
enum "ether-media-type-sfpp-dwdm-10gep-38-19" {
value 284;
}
enum "ether-media-type-sfpp-dwdm-10gep-37-79" {
value 285;
}
enum "ether-media-type-sfpp-dwdm-10gep-37-40" {
value 286;
}
enum "ether-media-type-sfpp-dwdm-10gep-37-00" {
value 287;
}
enum "ether-media-type-sfpp-dwdm-10gep-36-61" {
value 288;
}
enum "ether-media-type-sfpp-dwdm-10gep-36-22" {
value 289;
}
enum "ether-media-type-sfpp-dwdm-10gep-35-82" {
value 290;
}
enum "ether-media-type-sfpp-dwdm-10gep-35-43" {
value 291;
}
enum "ether-media-type-sfpp-dwdm-10gep-35-04" {
value 292;
}
enum "ether-media-type-sfpp-dwdm-10gep-34-64" {
value 293;
}
enum "ether-media-type-sfpp-dwdm-10gep-34-25" {
value 294;
}
enum "ether-media-type-sfpp-dwdm-10gep-33-86" {
value 295;
}
enum "ether-media-type-sfpp-dwdm-10gep-33-47" {
value 296;
}
enum "ether-media-type-sfpp-dwdm-10gep-33-07" {
value 297;
}
enum "ether-media-type-sfpp-dwdm-10gep-32-68" {
value 298;
}
enum "ether-media-type-sfpp-dwdm-10gep-32-29" {
value 299;
}
enum "ether-media-type-sfpp-dwdm-10gep-31-90" {
value 300;
}
enum "ether-media-type-sfpp-dwdm-10gep-31-51" {
value 301;
}
enum "ether-media-type-sfpp-dwdm-10gep-31-12" {
value 302;
}
enum "ether-media-type-sfpp-dwdm-10gep-30-72" {
value 303;
}
enum "ether-media-type-sfpp-dwdm-10gep-30-33" {
value 304;
}
enum "ether-media-type-ons-se-z1" {
value 305;
}
enum "ether-media-type-qsfp-100ge-lr" {
value 306;
}
enum "ether-media-type-qsfp-100ge-sr" {
value 307;
}
enum "ether-media-type-sfpp-10g-aoc1m" {
value 308;
}
enum "ether-media-type-sfpp-10g-aoc2m" {
value 309;
}
enum "ether-media-type-sfpp-10g-aoc3m" {
value 310;
}
enum "ether-media-type-sfpp-10g-aoc5m" {
value 311;
}
enum "ether-media-type-sfpp-10g-aoc7m" {
value 312;
}
enum "ether-media-type-sfpp-10g-aoc10m" {
value 313;
}
enum "ether-media-type-cpak-100ge-er4l" {
value 314;
}
enum "ether-media-type-qsfp-40ge-sr-bd" {
value 315;
}
enum "ether-media-type-qsfp-40ge-bd-rx" {
value 316;
}
enum "ether-media-type-tunable-dwdm-sfpp" {
value 317;
}
enum "ether-media-type-qsfp-40ge-sr-s" {
value 318;
}
enum "ether-media-type-qsfp-40ge-lr-s" {
value 319;
}
enum "ether-media-type-qsfp-h40ge-aoc1m" {
value 320;
}
enum "ether-media-type-qsfp-h40ge-aoc2m" {
value 321;
}
enum "ether-media-type-qsfp-h40ge-aoc3m" {
value 322;
}
enum "ether-media-type-qsfp-h40ge-aoc5m" {
value 323;
}
enum "ether-media-type-qsfp-h40ge-aoc7m" {
value 324;
}
enum "ether-media-type-qsfp-h40ge-aoc10m" {
value 325;
}
enum "ether-media-type-qsfp-h40ge-aoc15m" {
value 326;
}
enum "ether-media-type-qsfp-h40ge-aoc20m" {
value 327;
}
enum "ether-media-type-qsfp-40ge-csr4" {
value 328;
}
enum "ether-media-type-qsfp-40ge-sr4" {
value 329;
}
enum "ether-media-type-qsfp-40ge-lr4" {
value 330;
}
enum "ether-media-type-wsp-40ge-lr4l" {
value 331;
}
enum "ether-media-type-qsfp-h40ge-cu1m" {
value 332;
}
enum "ether-media-type-qsfp-h40ge-cu3m" {
value 333;
}
enum "ether-media-type-qsfp-h40ge-cu5m" {
value 334;
}
enum "ether-media-type-qsfp-h40ge-acu7m" {
value 335;
}
enum "ether-media-type-qsfp-h40ge-acu10m" {
value 336;
}
enum "ether-media-type-qsfp-100ge-psm4" {
value 337;
}
enum "ether-media-type-qsfp-100ge-cwdm4" {
value 338;
}
enum "ether-media-type-qsfp-100ge-cu1m" {
value 339;
}
enum "ether-media-type-qsfp-100ge-cu2m" {
value 340;
}
enum "ether-media-type-qsfp-100ge-cu3m" {
value 341;
}
enum "ether-media-type-qsfp-100ge-cu5m" {
value 342;
}
enum "ether-media-type-qsfp-100ge-aoc1m" {
value 343;
}
enum "ether-media-type-qsfp-100ge-aoc2m" {
value 344;
}
enum "ether-media-type-qsfp-100ge-aoc3m" {
value 345;
}
enum "ether-media-type-qsfp-100ge-aoc5m" {
value 346;
}
enum "ether-media-type-qsfp-100ge-aoc7m" {
value 347;
}
enum "ether-media-type-qsfp-100ge-aoc10m" {
value 348;
}
enum "ether-media-type-qsfp-100ge-aoc15m" {
value 349;
}
enum "ether-media-type-qsfp-100ge-aoc20m" {
value 350;
}
enum "ether-media-type-qsfp-100ge-aoc25m" {
value 351;
}
enum "ether-media-type-qsfp-100ge-aoc30m" {
value 352;
}
enum "ether-media-type-qsfp-100ge-sm-sr" {
value 353;
}
enum "ether-media-type-qsfp-100ge-csr4" {
value 354;
}
enum "ether-media-type-1000base-2bxd" {
value 355;
}
enum "ether-media-type-qsfp-h40ge-cudot5m" {
value 356;
}
enum "ether-media-type-qsfp-h40ge-cu2m" {
value 357;
}
enum "ether-media-type-qsfp-h40ge-cu4m" {
value 358;
}
enum "ether-media-type-virtual" {
value 359;
}
enum "ether-media-type-sfp-25ge-sr-s" {
value 360;
}
enum "ether-media-type-sfp-25ge-lr-s" {
value 361;
}
enum "ether-media-type-sfp-25ge-cu1m" {
value 362;
}
enum "ether-media-type-sfp-25ge-cu2m" {
value 363;
}
enum "ether-media-type-sfp-25ge-cu3m" {
value 364;
}
enum "ether-media-type-sfp-25ge-cu5m" {
value 365;
}
enum "ether-media-type-sfp-25ge-aoc1m" {
value 366;
}
enum "ether-media-type-sfp-25ge-aoc2m" {
value 367;
}
enum "ether-media-type-sfp-25ge-aoc3m" {
value 368;
}
enum "ether-media-type-sfp-25ge-aoc5m" {
value 369;
}
enum "ether-media-type-sfp-25ge-aoc7m" {
value 370;
}
enum "ether-media-type-sfp-25ge-aoc10m" {
value 371;
}
enum "ether-media-type-cvr-qsfp-sfp10g" {
value 372;
}
enum "ether-media-type-qsfp-h40ge-aoc25m" {
value 373;
}
enum "ether-media-type-qsfp-h40ge-aoc30m" {
value 374;
}
enum "ether-media-type-sfp-gpon" {
value 375;
}
enum "ether-media-type-sfp-25ge-csr" {
value 376;
}
enum "ether-media-type-10gbase-cu2m" {
value 377;
}
enum "ether-media-type-qsfp-100ge-er4l" {
value 378;
}
enum "ether-media-type-sfp-10ge-csr-s" {
value 379;
}
enum "ether-media-type-qsfp-40ge-csr-s" {
value 380;
}
enum "ether-media-type-qsfp28-4sfp25g-cu1m" {
value 381;
}
enum "ether-media-type-qsfp28-4sfp25g-cu2m" {
value 382;
}
enum "ether-media-type-qsfp28-4sfp25g-cu3m" {
value 383;
}
enum "ether-media-type-qsfp28-4sfp25g-cu5m" {
value 384;
}
enum "ether-media-type-10gbase-cu1-5m" {
value 385;
}
enum "ether-media-type-10gbase-cu2-5m" {
value 386;
}
enum "ether-media-type-qsfp-4x10g-lr-s" {
value 387;
}
enum "ether-media-type-qsfp-4x10g-aoc1m" {
value 388;
}
enum "ether-media-type-qsfp-4x10g-aoc2m" {
value 389;
}
enum "ether-media-type-qsfp-4x10g-aoc3m" {
value 390;
}
enum "ether-media-type-qsfp-4x10g-aoc5m" {
value 391;
}
enum "ether-media-type-qsfp-4x10g-aoc7m" {
value 392;
}
enum "ether-media-type-qsfp-4x10g-aoc10m" {
value 393;
}
enum "ether-media-type-qsfp-4x10g-aoc15m" {
value 394;
}
enum "ether-media-type-qsfp-4sfp10g-cu0-5m" {
value 395;
}
enum "ether-media-type-qsfp-4sfp10g-cu1m" {
value 396;
}
enum "ether-media-type-qsfp-4sfp10g-cu2m" {
value 397;
}
enum "ether-media-type-qsfp-4sfp10g-cu3m" {
value 398;
}
enum "ether-media-type-qsfp-4sfp10g-cu4m" {
value 399;
}
enum "ether-media-type-qsfp-4sfp10g-cu5m" {
value 400;
}
enum "ether-media-type-qsfp-4x10g-ac1m" {
value 401;
}
enum "ether-media-type-qsfp-4x10g-ac3m" {
value 402;
}
enum "ether-media-type-qsfp-4x10g-ac5m" {
value 403;
}
enum "ether-media-type-qsfp-4x10g-ac7m" {
value 404;
}
enum "ether-media-type-qsfp-4x10g-ac10m" {
value 405;
}
enum "ether-media-type-qsfp-100ge-sr4" {
value 406;
}
enum "ether-media-type-qsfp-100ge-lr4" {
value 407;
}
enum "ether-media-type-10gbase-cu0-5m" {
value 408;
}
enum "ether-media-type-10gbase-cu4m" {
value 409;
}
enum "ether-media-type-qsfp-40g-100g-srbd" {
value 410;
}
enum "ether-media-type-qsfp-h40ge-cu0-5m" {
value 411;
}
}
description
"Maximum possible values for media type";
}
typedef dot3-stats-versions {
type enumeration {
enum "not-supported" {
value 0;
description
"Dot 3 error counters are not supported";
}
enum "dot3-version-2" {
value 1;
description
"Version 2 dot 3 error counters";
}
}
description
"The version of dot 3 error counters.";
}
typedef serial-crc {
type enumeration {
enum "serial-crc32" {
value 0;
description
"32-bit Cyclic Redundancy Code";
}
enum "serial-crc16" {
value 1;
description
"16 bit Cyclic Redundancy Code";
}
}
description
"The Cyclic Redundancy Code type";
}
typedef subrate-speed {
type enumeration {
enum "dsx1-subrate-56kbps" {
value 0;
description
"56 kilobits per second subrate";
}
enum "dsx1-subrate-64kbps" {
value 1;
description
"64 kilobits per second subrate";
}
}
description
"The subrate on a serial interface";
}
typedef t1e1-loopback-mode {
type enumeration {
enum "t1e1-no-loopback" {
value 0;
description
"No loopback mode";
}
enum "t1e1-cli-local-loopback" {
value 1;
description
"Command line interface enforced local loopback";
}
enum "t1e1-line-cli-local-loopback" {
value 2;
description
"Command line interface enforced line local loopback";
}
enum "t1e1-payload-cli-local-loopback" {
value 3;
description
"Command line interface enforced payload local loopback";
}
enum "t1e1-local-line-loopback" {
value 4;
description
"Local line loopback";
}
enum "t1e1-local-payload-loopback" {
value 5;
description
"Local payload loopback";
}
enum "t1e1-local-ansi-fdl-remote-loopback" {
value 6;
description
"Line ANSI FDL remote loopback";
}
enum "t1e1-line-att-fdl-remote-loopback" {
value 7;
description
"Line ATT FDL remote loopback";
}
enum "t1e1-payload-ansi-fdl-remote-loopback" {
value 8;
description
"Payload ANSI FDL remote loopback";
}
enum "t1e1-payload-att-fdl-remote-loopback" {
value 9;
description
"Payload ATT FDL remote loopback";
}
enum "t1e1-line-iboc-remote-loopback" {
value 10;
description
"Line IBOC remote loopback";
}
enum "t1e1-line-ansi-fdl-local-loopback" {
value 11;
description
"Line ANSI FDL local loopback";
}
enum "t1e1-line-att-fdl-local-loopback" {
value 12;
description
"Line ATT FDL local loopback";
}
enum "t1e1-payload-ansi-fdl-local-loopback" {
value 13;
description
"Payload ANSI FDL local loopback";
}
enum "t1e1-payload-att-fdl-local-loopback" {
value 14;
description
"Payload ATT FDL local loopback";
}
enum "t1e1-line-iboc-local-loopback" {
value 15;
description
"Line IBOC local loopback";
}
}
description
"Loopback mode type";
}
typedef signal-status {
type enumeration {
enum "down" {
value 0;
description
"Signal is down";
}
enum "up" {
value 1;
description
"Signal is up";
}
}
description
"Synchronous serial interface signal status";
}
typedef idle-character-type {
type enumeration {
enum "flag" {
value 0;
description
"Send HDLC flag characters between packets";
}
enum "marks" {
value 1;
description
"Send mark characters between packets";
}
}
description
"Synchronous serial idle character";
}
grouping wred-class-counts {
description
"WRED class count types";
leaf wred-tx-pkts {
type uint64;
description
"Transmitted packets";
}
leaf wred-tx-bytes {
type uint64;
description
"Transmitted bytes";
}
leaf wred-tail-drop-pkts {
type uint64;
description
"Tail drop packets";
}
leaf wred-tail-drop-bytes {
type uint64;
description
"Tail drop bytes";
}
leaf wred-early-drop-pkts {
type uint64;
description
"Early drop packets";
}
leaf wred-early-drop-bytes {
type uint64;
description
"Early drop bytes";
}
}
grouping marking-dscp-stats {
description
"Statistics for set dscp";
leaf dscp {
type uint32;
description
"dscp marking";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-dscp-tunnel-stats {
description
"Statistics for set dscp tunnel";
leaf dscp-val {
type uint32;
description
"dscp value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-cos-stats {
description
"Statistics for set cos";
leaf cos-val {
type uint32;
description
"cos value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-cos-inner-stats {
description
"Statistics for set cos-inner";
leaf cos-inner-val {
type uint32;
description
"cos inner value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-discard-class-stats {
description
"Statistics for set discard-class";
leaf disc-class-val {
type uint32;
description
"discard-class value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-qos-grp-stats {
description
"Statistics for set qos-group";
leaf qos-grp-val {
type uint32;
description
"qos group value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-prec-stats {
description
"Statistics for set precedence";
leaf prec-val {
type uint32;
description
"precedence value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-prec-tunnel-stats {
description
"Statistics for set precedence tunnel";
leaf prec-val {
type uint32;
description
"precedence value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-mpls-exp-imp-stats {
description
"Statistics for set mpls exp imposition";
leaf mpls-exp-imp-val {
type uint32;
description
"mpls exp value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-mpls-exp-top-stats {
description
"Statistics for set mpls exp imposition";
leaf mpls-exp-top-val {
type uint32;
description
"mpls exp value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-fr-de-stats {
description
"Statistics for set fr-de";
leaf fr-de {
type boolean;
description
"fr de set or not";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-fr-fecn-becn-stats {
description
"Statistics for set fr-fecn-becn";
leaf fecn-becn-val {
type uint32;
description
"fecn becn value. qos:percent-value-1to100";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-atm-clp-stats {
description
"Statistics for set atm-clp";
leaf atm-clp-val {
type uint8;
description
"atm clp value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-vlan-inner-stats {
description
"Statistics for set vlan-inner";
leaf vlan-inner-val {
type uint32;
description
"vlan value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-dei-stats {
description
"Statistics for set dei";
leaf dei-imp-value {
type uint32;
description
"dei value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-dei-imp-stats {
description
"Statistics for set dei-imposition";
leaf dei-imp-value {
type uint32;
description
"dei value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-srp-priority-stats {
description
"Statistics for set srp-priority";
leaf srp-priority-value {
type uint8;
description
"srp priority value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-wlan-user-priority-stats {
description
"Statistics for set wlan-user-priority";
leaf wlan-user-priority-value {
type uint8;
description
"wlan user priority value";
}
leaf marked-pkts {
type uint64;
description
"Number of packets been marked";
}
}
grouping marking-stats {
description
"Marking statistics";
container marking-dscp-stats-val {
description
"Statistics for set dscp";
uses interfaces-ios-xe-oper:marking-dscp-stats;
}
container marking-dscp-tunnel-stats-val {
description
"Statistics for set dscp tunnel";
uses interfaces-ios-xe-oper:marking-dscp-tunnel-stats;
}
container marking-cos-stats-val {
description
"Statistics for set cos";
uses interfaces-ios-xe-oper:marking-cos-stats;
}
container marking-cos-inner-stats-val {
description
"Statistics for set cos-inner";
uses interfaces-ios-xe-oper:marking-cos-inner-stats;
}
container marking-discard-class-stats-val {
description
"Statistics for set discard-class";
uses interfaces-ios-xe-oper:marking-discard-class-stats;
}
container marking-qos-grp-stats-val {
description
"Statistics for set qos-group";
uses interfaces-ios-xe-oper:marking-qos-grp-stats;
}
container marking-prec-stats-val {
description
"Statistics for set precedence";
uses interfaces-ios-xe-oper:marking-prec-stats;
}
container marking-prec-tunnel-stats-val {
description
"Statistics for set precedence tunnel";
uses interfaces-ios-xe-oper:marking-prec-tunnel-stats;
}
container marking-mpls-exp-imp-stats-val {
description
"Statistics for set mpls exp imposition";
uses interfaces-ios-xe-oper:marking-mpls-exp-imp-stats;
}
container marking-mpls-exp-top-stats-val {
description
"Statistics for set mpls exp topmost";
uses interfaces-ios-xe-oper:marking-mpls-exp-top-stats;
}
container marking-fr-de-stats-val {
description
"Statistics for set fr-de";
uses interfaces-ios-xe-oper:marking-fr-de-stats;
}
container marking-fr-fecn-becn-stats-val {
description
"Statistics for set fr-fecn-becn";
uses interfaces-ios-xe-oper:marking-fr-fecn-becn-stats;
}
container marking-atm-clp-stats-val {
description
"Statistics for set atm-clp";
uses interfaces-ios-xe-oper:marking-atm-clp-stats;
}
container marking-vlan-inner-stats-val {
description
"Statistics for set vlan-inner";
uses interfaces-ios-xe-oper:marking-vlan-inner-stats;
}
container marking-dei-stats-val {
description
"Statistics for set dei";
uses interfaces-ios-xe-oper:marking-dei-stats;
}
container marking-dei-imp-stats-val {
description
"Statistics for set dei-imposition";
uses interfaces-ios-xe-oper:marking-dei-imp-stats;
}
container marking-srp-priority-stats-val {
description
"Statistics for set srp-priority";
uses interfaces-ios-xe-oper:marking-srp-priority-stats;
}
container marking-wlan-user-priority-stats-val {
description
"Statistics for set wlan-user-priority";
uses interfaces-ios-xe-oper:marking-wlan-user-priority-stats;
}
}
grouping cos-key {
description
"COS key type";
leaf cos-min {
type uint32;
description
"Min COS value";
}
leaf cos-max {
type uint32;
description
"Max COS value";
}
}
grouping disc-class-key {
description
"Discard class key type";
leaf disc-class-min {
type uint32;
description
"Minimum value for discard class in the range";
}
leaf disc-class-max {
type uint32;
description
"Maximum value for discard class in the range";
}
}
grouping dscp-key {
description
"DSCP key type";
leaf dscp-min {
type uint32;
description
"Minimum of dscp range";
}
leaf dscp-max {
type uint32;
description
"Maximum of dscp range";
}
}
grouping qos-grp-key {
description
"QoS group key type";
leaf qos-group-min {
type uint32;
description
"Specifies the minimum value range from 0 to used to identify a QoS group value";
}
leaf qos-group-max {
type uint32;
description
"Specifies the maximum value range from 0 to used to identify a QoS group value";
}
}
grouping mpls-exp-key {
description
"MPLS EXP key type";
leaf exp-min {
type uint32;
description
"The minimum EXP field value to be used as match criteria. Any number from 0 to 7";
}
leaf exp-max {
type uint32;
description
"The maximum EXP field value to be used as match criteria. Any number from 0 to 7";
}
}
grouping prec-key {
description
"Precedence key type";
leaf prec-min {
type uint32;
description
"Precedence min";
}
leaf prec-max {
type uint32;
description
"Precedence max";
}
}
grouping dei-key {
description
"DEI key type";
leaf dei-min {
type uint32;
description
"Dei min";
}
leaf dei-max {
type uint32;
description
"Dei max";
}
}
grouping clp-key {
description
"CLP key type";
leaf clp-val {
type uint32;
description
"Composed by multiple atm clp values";
}
}
grouping subclass-list {
description
"Subclass data";
leaf match-type {
type interfaces-ios-xe-oper:qos-match-type;
description
"Subclass match type";
}
list cos-counters {
key "cos-min cos-max";
description
"Counters for sub-class matching a range of
Class-of-Service (COS) value (and, optionally, additional COS range";
uses interfaces-ios-xe-oper:cos-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container cos-default {
description
"statistics for cos default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
list dscp-counters {
key "dscp-min dscp-max";
description
"List for statistics based on dscp value range";
uses interfaces-ios-xe-oper:dscp-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container dscp-default {
description
"Statistics for dscp default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
list discard-class-counters {
key "disc-class-min disc-class-max";
description
"Composed multiple discard class ranges";
uses interfaces-ios-xe-oper:disc-class-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container disc-class-default {
description
"Statistics for discard class default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
list precedence-counters {
key "prec-min prec-max";
description
"List for statistics based on precedence value range";
uses interfaces-ios-xe-oper:prec-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container prec-default {
description
"Precedence default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
list mpls-exp-counters {
key "exp-min exp-max";
description
"List for statistics based on mpls exp value range";
uses interfaces-ios-xe-oper:mpls-exp-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container mpls-exp-default {
description
"Statistics for mpls-exp default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
list dei-counters {
key "dei-min dei-max";
description
"Composed by multiple dei ranges";
uses interfaces-ios-xe-oper:dei-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container dei-counts-default {
description
"Statistics for dei default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
list clp-counters {
key "clp-val";
description
"Statistics for each value range for a specific subclass type";
uses interfaces-ios-xe-oper:clp-key;
uses interfaces-ios-xe-oper:wred-class-counts;
}
container clp-default {
description
"Statistic for atm clp default";
uses interfaces-ios-xe-oper:wred-class-counts;
}
}
grouping classifier-entry-statistics {
description
"Classifier entry statistics";
leaf classified-pkts {
type uint64;
description
"Number of total packets which filtered to the classifier-entry";
}
leaf classified-bytes {
type uint64;
description
"Number of total bytes which filtered to the classifier-entry";
}
leaf classified-rate {
type uint64;
description
"Rate of average data flow through the classifier-entry";
}
}
grouping diffserv-target-classifier-statistics-key {
description
"Diffserv classifier statistics key";
leaf classifier-entry-name {
type string;
description
"Classifier Entry Name";
}
leaf parent-path {
type string;
description
"Path of the Classifier Entry in a hierarchical policy";
}
}
grouping wred-stats {
description
"WRED counters";
leaf early-drop-pkts {
type uint64;
description
"Early drop packets";
}
leaf early-drop-bytes {
type uint64;
description
"Early drop bytes";
}
leaf mean-queue-depth {
type uint16;
description
"Current mean queue depth";
}
leaf transmitted-pkts {
type uint64;
description
"Transmitted packets";
}
leaf transmitted-bytes {
type uint64;
description
"Transmitted bytes";
}
leaf tail-drop-pkts {
type uint64;
description
"Total number of packets dropped";
}
leaf tail-drop-bytes {
type uint64;
description
"Total number of bytes dropped";
}
leaf drop-pkts-flow {
type uint64;
description
"Total number of packets dropped";
}
leaf drop-pkts-no-buffer {
type uint64;
description
"Number of packets dropped due to buffers being
unavailable system-wide or at the associated interface.
This is a sub-set of drop-pkts";
}
leaf queue-peak-size-pkts {
type uint64;
description
"Queue max que depth Packets";
}
leaf queue-peak-size-bytes {
type uint64;
description
"Queue max que depth Bytes";
}
leaf bandwidth-exceed-drops {
type uint64;
description
"Priority stats. Bandwidth exceed drops";
}
}
grouping cac-stats {
description
"CAC statistics";
leaf num-admitted-flows {
type uint32;
description
"Number of admitted flows";
}
leaf num-non-admitted-flows {
type uint32;
description
"Number of non-admitted flows";
}
}
grouping queuing-statistics {
description
"Queue related statistics";
leaf output-pkts {
type uint64;
description
"Number of packets transmitted from queue";
}
leaf output-bytes {
type uint64;
description
"Number of bytes transmitted from queue";
}
leaf queue-size-pkts {
type uint64;
description
"Number of packets currently buffered ";
}
leaf queue-size-bytes {
type uint64;
description
"Number of bytes currently buffered";
}
leaf drop-pkts {
type uint64;
description
"Total number of packets dropped";
}
leaf drop-bytes {
type uint64;
description
"Total number of bytes dropped";
}
container wred-stats {
description
"WRED Counters";
uses interfaces-ios-xe-oper:wred-stats;
}
container cac-stats {
description
"CAC statistics";
uses interfaces-ios-xe-oper:cac-stats;
}
}
grouping meter-statistics {
description
"Metering Counters";
leaf meter-id {
type uint16;
description
"Meter Identifier";
}
leaf meter-succeed-pkts {
type uint64;
description
"Number of packets which succeed the meter";
}
leaf meter-succeed-bytes {
type uint64;
description
"Bytes of packets which succeed the meter";
}
leaf meter-failed-pkts {
type uint64;
description
"Number of packets which failed the meter";
}
leaf meter-failed-bytes {
type uint64;
description
"Bytes of packets which failed the meter";
}
}
grouping diffserv-target-classifier-statistics {
description
"Diffserv classifier statistics";
container classifier-entry-stats {
description
"Classifier Counters";
uses interfaces-ios-xe-oper:classifier-entry-statistics;
}
list meter-stats {
key "meter-id";
description
"Meter statistics";
uses interfaces-ios-xe-oper:meter-statistics;
}
container queuing-stats {
description
"Queuing Counters";
uses interfaces-ios-xe-oper:queuing-statistics;
}
list subclass-list {
key "match-type";
description
"List of statistics for random-detect
based on subclass type and value pair
Technically these are a field in the queuing
statistics -> wred statistics, but GREEN EI
does not allow that nesting structure";
uses interfaces-ios-xe-oper:subclass-list;
}
container marking-stats {
description
"Statistics for marking actions";
uses interfaces-ios-xe-oper:marking-stats;
}
}
grouping diffserv-target-entry-key {
description
"Key to the diffserv";
leaf direction {
type interfaces-ios-xe-oper:qos-direction;
description
"Direction fo the traffic flow either inbound or outbound";
}
leaf policy-name {
type string;
description
"Policy entry name";
}
}
grouping agg-priority-stats {
description
"Type for counters in aggregate priority";
leaf output-pkts {
type uint64;
description
"Number of packets transmitted from queue";
}
leaf output-bytes {
type uint64;
description
"Number of bytes transmitted from queue";
}
leaf queue-size-pkts {
type uint64;
description
"Number of packets currently buffered";
}
leaf queue-size-bytes {
type uint64;
description
"Number of bytes currently buffered";
}
leaf drop-pkts {
type uint64;
description
"Total number of packets dropped";
}
leaf drop-bytes {
type uint64;
description
"Total number of bytes dropped";
}
leaf drop-pkts-flow {
type uint64;
description
"Number of packets that were dropped by
flow-based fair-queuing (fair-queue).
This is a sub-set of drop-pkts";
}
leaf drop-pkts-no-buffer {
type uint64;
description
"Number of packets dropped due to buffers being
unavailable system-wide or at the associated interface.
This is a sub-set of drop-pkts";
}
}
grouping threshold {
description
"Threshold Parameters";
leaf bytes {
type uint64;
description
"Threshold bytes";
}
leaf thresh-size-metric {
type uint32;
description
"Threshold size unit";
}
leaf unit-val {
type interfaces-ios-xe-oper:thresh-unit;
description
"Threshold size basic units";
}
leaf threshold-interval {
type uint64;
description
"Threshold interval";
}
leaf thresh-interval-metric {
type uint32;
description
"Threshold units metric";
}
leaf interval-unit-val {
type interfaces-ios-xe-oper:thresh-unit;
description
"Threshold interval basic units";
}
}
grouping priority-oper-list {
description
"Priority based statistics";
leaf priority-level {
type uint16;
description
"Priority Level, 0 means no priority level set";
}
container agg-priority-stats {
description
"Counters in aggregate priority";
uses interfaces-ios-xe-oper:agg-priority-stats;
}
container qlimit-default-thresh {
description
"queue limit default threshold";
uses interfaces-ios-xe-oper:threshold;
}
list qlimit-cos-thresh-list {
key "cos-min cos-max";
description
"cos-based queue limit data";
uses interfaces-ios-xe-oper:cos-key;
uses interfaces-ios-xe-oper:threshold;
}
list qlimit-disc-class-thresh-list {
key "disc-class-min disc-class-max";
description
"discard-class-based queue limit data";
uses interfaces-ios-xe-oper:disc-class-key;
uses interfaces-ios-xe-oper:threshold;
}
list qlimit-qos-grp-thresh-list {
key "qos-group-min qos-group-max";
description
"qos-group-based queue limit data";
uses interfaces-ios-xe-oper:qos-grp-key;
uses interfaces-ios-xe-oper:threshold;
}
list qlimit-mpls-exp-thresh-list {
key "exp-min exp-max";
description
"mpls-exp-based queue limit data";
uses interfaces-ios-xe-oper:mpls-exp-key;
uses interfaces-ios-xe-oper:threshold;
}
list qlimit-dscp-thresh-list {
key "dscp-min dscp-max";
description
"queue limit per dscp range";
uses interfaces-ios-xe-oper:dscp-key;
uses interfaces-ios-xe-oper:threshold;
}
}
grouping diffserv-target-entry {
description
"Diffserv target data";
list diffserv-target-classifier-stats {
key "classifier-entry-name parent-path";
description
"Statistics for each Classifier Entry in a Policy";
uses interfaces-ios-xe-oper:diffserv-target-classifier-statistics-key;
uses interfaces-ios-xe-oper:diffserv-target-classifier-statistics;
}
list priority-oper-list {
key "priority-level";
description
"Statistics for aggregate priority per policy instance";
uses interfaces-ios-xe-oper:priority-oper-list;
}
}
grouping lag-aggregate-state {
description
"Operational state variables for logical
aggregate / LAG interfaces";
leaf aggregate-id {
type string;
description
"Specify the logical aggregate interface to which
this id belongs";
}
leaf lag-type {
type interfaces-ios-xe-oper:aggregation-type;
description
"Type to define the lag-type, i.e., how the LAG is
defined and managed";
}
leaf min-links {
type uint16;
description
"Specifies the minimum number of member
interfaces that must be active for the aggregate interface
to be available";
}
leaf lag-speed {
type uint32;
description
"Reports effective speed of the aggregate interface,
based on speed of active member interfaces";
}
leaf-list members {
type string;
ordered-by user;
description
"List of current member interfaces for the aggregate,
expressed as references to existing interfaces";
}
}
grouping protocol-statistics {
description
"Protocol specific statistics";
leaf in-pkts {
type uint64;
description
"The total number of packets received for the specified
address family, including those received in error";
}
leaf in-octets {
type uint64;
description
"The total number of octets received in input packets
for the specified address family, including those received
in error.";
}
leaf in-error-pkts {
type uint64;
description
"Number of packets discarded due to errors for the
specified address family, including errors in the
header, no route found to the destination, invalid
address, unknown protocol, etc.";
}
leaf in-forwarded-pkts {
type uint64;
description
"The number of input packets for which the device was not
their final destination and for which the device
attempted to find a route to forward them to that final
destination.";
}
leaf in-forwarded-octets {
type uint64;
description
"The number of octets received in input packets
for the specified address family for which the device was
not their final destination and for which the
device attempted to find a route to forward them to that
final destination.";
}
leaf in-discarded-pkts {
type uint64;
description
"The number of input IP packets for the
specified address family, for which no problems were
encountered to prevent their continued processing, but
were discarded (e.g., for lack of buffer space).";
}
leaf out-pkts {
type uint64;
description
"The total number of IP packets for the
specified address family that the device supplied
to the lower layers for transmission. This includes
packets generated locally and those forwarded by the
device.";
}
leaf out-octets {
type uint64;
description
"The total number of octets in IP packets for the
specified address family that the device
supplied to the lower layers for transmission. This
includes packets generated locally and those forwarded by
the device.";
}
leaf out-error-pkts {
type uint64;
description
"Number of IP packets for the specified address family
locally generated and discarded due to errors, including
no route found to the IP destination.";
}
leaf out-forwarded-pkts {
type uint64;
description
"The number of packets for which this entity was not their
final IP destination and for which it was successful in
finding a path to their final destination.text";
}
leaf out-forwarded-octets {
type uint64;
description
"The number of octets in packets for which this entity was
not their final IP destination and for which it was
successful in finding a path to their final destination.";
}
leaf out-discarded-pkts {
type uint64;
description
"The number of output IP packets for the
specified address family for which no problem was
encountered to prevent their transmission to their
destination, but were discarded (e.g., for lack of
buffer space).";
}
}
grouping intf-statistics {
description
"Interface statistics";
leaf discontinuity-time {
type yang:date-and-time;
description
"The time on the most recent occasion at which any one or
more of this interface's counters suffered a
discontinuity. If no such discontinuities have occurred
since the last re-initialization of the local management
subsystem, then this node contains the time the local
management subsystem re-initialized itself";
}
leaf in-octets {
type uint64;
description
"The total number of octets received on the interface,
including framing characters.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of discontinuity-time";
}
leaf in-unicast-pkts {
type uint64;
description
"The number of packets, delivered by this sub-layer to a
higher (sub-)layer, that were not addressed to a
multicast or broadcast address at this sub-layer.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'.";
}
leaf in-broadcast-pkts {
type uint64;
description
"The number of packets, delivered by this sub-layer to a
higher (sub-)layer, that were addressed to a broadcast
address at this sub-layer.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf in-multicast-pkts {
type uint64;
description
"The number of packets, delivered by this sub-layer to a
higher (sub-)layer, that were addressed to a multicast
address at this sub-layer. For a MAC-layer protocol,
this includes both Group and Functional addresses.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf in-discards {
type uint32;
description
"The number of inbound packets that were chosen to be
discarded even though no errors had been detected to
prevent their being deliverable to a higher-layer
protocol. One possible reason for discarding such a
packet could be to free up buffer space.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf in-errors {
type uint32;
description
"For packet-oriented interfaces, the number of inbound
packets that contained errors preventing them from being
deliverable to a higher-layer protocol. For character-
oriented or fixed-length interfaces, the number of
inbound transmission units that contained errors
preventing them from being deliverable to a higher-layer protocol.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf in-unknown-protos {
type uint32;
description
"For packet-oriented interfaces, the number of packets
received via the interface that were discarded because
of an unknown or unsupported protocol. For
character-oriented or fixed-length interfaces that
support protocol multiplexing, the number of transmission
units received via the interface that were discarded because
of an unknown or unsupported protocol.
For any interface that does not support protocol
multiplexing, this counter is not present.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of 'discontinuity-time'";
}
leaf out-octets {
type uint32;
description
"The total number of octets transmitted out of the
interface, including framing characters.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf out-unicast-pkts {
type uint64;
description
"The total number of packets that higher-level protocols
requested be transmitted, and that were not addressed
to a multicast or broadcast address at this sub-layer,
including those that were discarded or not sent.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of 'discontinuity-time'";
}
leaf out-broadcast-pkts {
type uint64;
description
"The total number of packets that higher-level protocols
requested be transmitted, and that were addressed to a
broadcast address at this sub-layer, including those
that were discarded or not sent.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf out-multicast-pkts {
type uint64;
description
"The total number of packets that higher-level protocols
requested be transmitted, and that were addressed to a
multicast address at this sub-layer, including those
that were discarded or not sent. For a MAC-layer
protocol, this includes both Group and Functional addresses.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf out-discards {
type uint64;
description
"The number of outbound packets that were chosen to be
discarded even though no errors had been detected to
prevent their being transmitted. One possible reason
for discarding such a packet could be to free up buffer space.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf out-errors {
type uint64;
description
"For packet-oriented interfaces, the number of outbound
packets that could not be transmitted because of errors.
For character-oriented or fixed-length interfaces, the
number of outbound transmission units that could not be
transmitted because of errors.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf rx-pps {
type uint64;
description
"The receive packet per second rate on this interface";
}
leaf rx-kbps {
type uint64;
description
"The receive kilobits per second rate on this interface";
}
leaf tx-pps {
type uint64;
description
"The transmit packet per second rate on this interface";
}
leaf tx-kbps {
type uint64;
description
"The transmit kilobits per second rate on this interface";
}
leaf num-flaps {
type uint64;
description
"The number of times the interface state
transitioned between up and down";
}
leaf in-crc-errors {
type uint64;
description
"Number of receive error events due to FCS/CRC check
failure";
}
leaf in-discards-64 {
type uint64;
description
"The number of inbound packets that were chosen to be
discarded even though no errors had been detected to
prevent their being deliverable to a higher-layer
protocol. One possible reason for discarding such a
packet could be to free up buffer space.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf in-errors-64 {
type uint64;
description
"For packet-oriented interfaces, the number of inbound
packets that contained errors preventing them from being
deliverable to a higher-layer protocol. For character-
oriented or fixed-length interfaces, the number of
inbound transmission units that contained errors
preventing them from being deliverable to a higher-layer protocol.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
leaf in-unknown-protos-64 {
type uint64;
description
"For packet-oriented interfaces, the number of packets
received via the interface that were discarded because
of an unknown or unsupported protocol. For
character-oriented or fixed-length interfaces that
support protocol multiplexing, the number of transmission
units received via the interface that were discarded because
of an unknown or unsupported protocol.
For any interface that does not support protocol
multiplexing, this counter is not present.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of 'discontinuity-time'";
}
leaf out-octets-64 {
type uint64;
description
"The total number of octets transmitted out of the
interface, including framing characters.
Discontinuities in the value of this counter can occur
at re-initialization of the management system, and at
other times as indicated by the value of
'discontinuity-time'";
}
}
grouping ethernet-state {
description
"Ethernet State";
leaf negotiated-duplex-mode {
type interfaces-ios-xe-oper:ether-duplex;
description
"When auto-negotiate is set to TRUE, and the
interface has completed auto-negotiation with the
remote peer, this value shows the duplex mode that
has been negotiated.";
}
leaf negotiated-port-speed {
type interfaces-ios-xe-oper:ether-speed;
description
"When auto-negotiate is set to TRUE, and
the interface has completed auto-negotiation with
the remote peer, this value shows the interface
speed that has been negotiated.";
}
leaf auto-negotiate {
type boolean;
description
"Set to TRUE to request the interface to
auto-negotiate transmission parameters with its
peer interface. When set to FALSE, the transmission
parameters are specified manually.";
}
leaf enable-flow-control {
type boolean;
description
"Enable or disable flow control for this
interface. Ethernet flow control is a mechanism by
which a receiver may send PAUSE frames to a sender
to stop transmission for a specified time.
This setting should override auto-negotiated flow
control settings. If left unspecified, and
auto-negotiate is TRUE, flow control mode is
negotiated with the peer interface.";
}
leaf media-type {
type interfaces-ios-xe-oper:media-type-class;
description
"Type of media attached to the interface";
}
}
grouping dot3-v2-error-counters {
description
"dot 3 version 2 error counters";
leaf dot3-alignment-errors {
type uint64;
description
"A count of frames received that are not an integral number of
octets in length and do not pass the FCS check.";
}
leaf dot3-fcs-errors {
type uint64;
description
"A count of frames received that are an integral number of
octets in length and do not pass the FCS check.";
}
leaf dot3-single-collision-frames {
type uint64;
description
"A count of frames that are involved in a single
collision, and are subsequently transmitted
successfully.";
}
leaf dot3-multiple-collision-frames {
type uint64;
description
"A count of frames that are involved in more
than one collision and are subsequently
transmitted successfully.";
}
leaf dot3-sqe-test-errors {
type uint64;
description
"A count of times that the signal quality error
test errors is received on a particular interface.";
}
leaf dot3-deferred-transmissions {
type uint64;
description
"A count of frames for which the first
transmission attempt on a particular interface
is delayed because the medium is busy.";
}
leaf dot3-late-collisions {
type uint64;
description
"The number of times that a collision is
detected on a particular interface later than
one slot time into the transmission of a packet.";
}
leaf dot3-excessive-collisions {
type uint64;
description
"A count of frames for which transmission on a
particular interface fails due to excessive
collisions.";
}
leaf dot3-internal-mac-transmit-errors {
type uint64;
description
"A count of frames for which transmission on a
particular interface fails due to an internal
MAC sublayer transmit error.";
}
leaf dot3-carrier-sense-errors {
type uint64;
description
"The number of times that the carrier sense
condition was lost or never asserted when
attempting to transmit a frame on a particular
interface.";
}
leaf dot3-frame-too-longs {
type uint64;
description
"A count of frames received on a particular
interface that exceed the maximum permitted
frame size.";
}
leaf dot3-internal-mac-receive-errors {
type uint64;
description
"A count of frames for which reception on a
particular interface fails due to an internal
MAC sublayer receive error.";
}
leaf dot3-symbol-errors {
type uint64;
description
"For an interface operating at 100 Mb/s, the
number of times there was an invalid data symbol
when a valid carrier was present.";
}
leaf dot3-duplex-status {
type uint64;
description
"The current mode of operation of the MAC
entity. 'unknown' indicates that the current
duplex mode could not be determined.
The possible value as defined by RFC 3635 are:
unknown 1
half duplex 2
full duplex 3";
}
leaf dot3-hc-alignment-errors {
type uint64;
description
"A count of frames received that are not an integral number of
octets in length and do not pass the FCS check on interfaces
operating at 10 Gb/s or faster.";
}
leaf dot3-hc-inpause-frames {
type uint64;
description
"MAC layer PAUSE frames received on the interface";
}
leaf dot3-hc-outpause-frames {
type uint64;
description
"MAC layer PAUSE frames sent on the interface";
}
leaf dot3-hc-fcs-errors {
type uint64;
description
"A count of frames received that are an integral number of
octets in length and do not pass the FCS check on interfaces
operating at 10 Gb/s or faster.";
}
leaf dot3-hc-frame-too-longs {
type uint64;
description
"A count of frames received on a particular
interface that exceed the maximum permitted
frame size on interfaces operating at 10 Gb/s or faster.";
}
leaf dot3-hc-internal-mac-transmit-errors {
type uint64;
description
"A count of frames for which transmission on a
particular interface fails due to an internal
MAC sublayer transmit error on interfaces
operating at 10 Gb/s or faster.";
}
leaf dot3-hc-internal-mac-receive-errors {
type uint64;
description
"A count of frames for which reception on a
particular interface fails due to an internal
MAC sublayer receive error on interfaces
operating at 10 Gb/s or faster.";
}
leaf dot3-hc-symbol-errors {
type uint64;
description
"For an interface operating at 10 Gb/s or higher, the
number of times the receiving media is non-idle
(a carrier event) for a period of time equal to
or greater than minimum frame size, and during which
there was at least one occurrence of an event
that causes the PHY to indicate 'Receive Error'";
}
}
grouping dot3-counters {
description
"Dot 3 error counters";
leaf dot3-stats-version {
type interfaces-ios-xe-oper:dot3-stats-versions;
description
"Version of dot 3 error counters";
}
container dot3-error-counters-v2 {
description
"The Ethernet dot 3 version 2 error counters";
uses interfaces-ios-xe-oper:dot3-v2-error-counters;
}
}
grouping ethernet-statistics {
description
"Ethernet statistics";
leaf in-mac-control-frames {
type uint64;
description
"MAC layer control frames received on the interface";
}
leaf in-mac-pause-frames {
type uint64;
description
"MAC layer PAUSE frames received on the interface";
}
leaf in-oversize-frames {
type uint64;
description
"Number of oversize frames received on the interface";
}
leaf in-jabber-frames {
type uint64;
description
"Number of jabber frames received on the
interface. Jabber frames are typically defined as oversize
frames which also have a bad CRC. Implementations may use
slightly different definitions of what constitutes a jabber
frame. Often indicative of a NIC hardware problem.";
}
leaf in-fragment-frames {
type uint64;
description
"Number of fragment frames received on the interface.";
}
leaf in-8021q-frames {
type uint64;
description
"Number of 802.1q tagged frames received on the interface";
}
leaf out-mac-control-frames {
type uint64;
description
"MAC layer control frames sent on the interface";
}
leaf out-mac-pause-frames {
type uint64;
description
"MAC layer PAUSE frames sent on the interface";
}
leaf out-8021q-frames {
type uint64;
description
"Number of 802.1q tagged frames sent on the interface";
}
leaf dot3-counters-supported {
type empty;
description
"When present, the dot 3 version 2 error counters
are valid";
}
container dot3-counters {
when "boolean(../dot3-counters-supported)";
description
"dot 3 error counters";
uses interfaces-ios-xe-oper:dot3-counters;
}
}
grouping t1e1-serial-state {
description
"The T1/E1 serial state";
leaf crc-type {
type interfaces-ios-xe-oper:serial-crc;
description
"Cyclic Redundancy Code type configured on the interface";
}
leaf loopback {
type interfaces-ios-xe-oper:t1e1-loopback-mode;
description
"Loopback mode the interface is operating in";
}
leaf keeplive {
type uint32;
description
"Keep alive interval in seconds";
}
leaf timeslot {
type uint32;
description
"Time slots bitmap occupied by this serial interface";
}
leaf subrate {
type interfaces-ios-xe-oper:subrate-speed;
description
"Subrate operating per slot";
}
}
grouping t1e1-serial-stats {
description
"The serial specific statistics";
leaf in-abort-clock-error {
type uint32;
description
"Number of receive abort packets due to clock slides";
}
}
grouping dce-state {
description
"Synchronous serial DCE mode state";
leaf dce-terminal-timing-enable {
type boolean;
description
"Enable DCE terminal timing";
}
leaf ignore-dtr {
type boolean;
description
"Ignore DTR signal";
}
leaf serial-clock-rate {
type uint32;
units "bps";
description
"Serial clock rate in bps";
}
}
grouping dte-state {
description
"Synchronous serial DTE mode state";
leaf tx-invert-clk {
type boolean;
description
"Invert transmit clock";
}
leaf ignore-dcd {
type boolean;
description
"Ignore DCD signal";
}
leaf rx-clockrate {
type uint32;
units "bps";
description
"Received clock rate";
}
leaf rx-clock-threshold {
type uint32;
units "bps";
description
"Received clock rate threshold limit";
}
}
grouping sync-serial-state {
description
"The synchronous serial state";
leaf carrier-delay {
type uint32;
units "milli-seconds";
description
"Delay for interface transitions";
}
leaf dtr-pulse-time {
type uint32;
units "milli-seconds";
description
"DTR pulse time on reset";
}
leaf restart-delay {
type uint32;
units "milli-seconds";
description
"Serial interface restart-delay";
}
leaf cable-type {
type string;
description
"Cable type attached on the interface";
}
leaf loopback {
type boolean;
description
"Whether the interface is in loopback mode";
}
leaf nrzi-encoding {
type boolean;
description
"Enable use of NRZI encoding";
}
leaf idle-character {
type interfaces-ios-xe-oper:idle-character-type;
description
"Idle character type";
}
leaf rts-signal {
type interfaces-ios-xe-oper:signal-status;
description
"RTS signal state";
}
leaf cts-signal {
type interfaces-ios-xe-oper:signal-status;
description
"CTS signal state";
}
leaf dtr-signal {
type interfaces-ios-xe-oper:signal-status;
description
"DTR signal state";
}
leaf dcd-signal {
type interfaces-ios-xe-oper:signal-status;
description
"DCD signal state";
}
leaf dsr-signal {
type interfaces-ios-xe-oper:signal-status;
description
"DSR signal state";
}
container dce-mode-state {
description
"Special state in DCE mode";
uses interfaces-ios-xe-oper:dce-state;
}
container dte-mode-state {
description
"Special state in DTE mode";
uses interfaces-ios-xe-oper:dte-state;
}
}
grouping interface-state {
description
"Interface state details";
leaf name {
type string;
description
"The name of the interface.
A server implementation MAY map this leaf to the ifName
MIB object. Such an implementation needs to use some
mechanism to handle the differences in size and characters
allowed between this leaf and ifName. The definition of
such a mechanism is outside the scope of this document";
}
leaf interface-type {
type interfaces-ios-xe-oper:ietf-intf-type;
description
"When an interface entry is created, a server MAY
initialize the type leaf with a valid value, e.g., if it
is possible to derive the type from the name of the interface.
If a client tries to set the type of an interface to a
value that can never be used by the system, e.g., if the
type is not supported or if the type does not match the
name of the interface, the server MUST reject the request.
A NETCONF server MUST reply with an rpc-error with the
error-tag 'invalid-value' in this case";
}
leaf admin-status {
type interfaces-ios-xe-oper:intf-state;
description
"The desired state of the interface.
This leaf has the same read semantics as ifAdminStatus";
}
leaf oper-status {
type interfaces-ios-xe-oper:oper-state;
description
"The current operational state of the interface.
This leaf has the same semantics as ifOperStatus";
}
leaf last-change {
type yang:date-and-time;
description
"The time the interface entered its current operational
state. If the current state was entered prior to the
last re-initialization of the local network management
subsystem, then this node is not present";
}
leaf if-index {
type int32;
description
"The ifIndex value for the ifEntry represented by this interface";
}
leaf phys-address {
type yang:mac-address;
description
"The interface's address at its protocol sub-layer. For
example, for an 802.x interface, this object normally
contains a Media Access Control (MAC) address. The
interface's media-specific modules must define the bit
and byte ordering and the format of the value of this
object. For interfaces that do not have such an address
(e.g., a serial line), this node is not present";
}
leaf-list higher-layer-if {
type string;
ordered-by user;
description
"A list of references to interfaces layered on top of this interface";
}
leaf-list lower-layer-if {
type string;
ordered-by user;
description
"A list of references to interfaces layered underneath this interface";
}
leaf speed {
type uint64;
description
"An estimate of the interface's current bandwidth in bits
per second. For interfaces that do not vary in
bandwidth or for those where no accurate estimation can
be made, this node should contain the nominal bandwidth.
For interfaces that have no concept of bandwidth, this
node is not present";
}
container statistics {
description
"A collection of interface-related statistics objects";
uses interfaces-ios-xe-oper:intf-statistics;
}
list diffserv-info {
key "direction policy-name";
description
"diffserv related details";
uses interfaces-ios-xe-oper:diffserv-target-entry-key;
uses interfaces-ios-xe-oper:diffserv-target-entry;
}
leaf vrf {
type string;
description
"VRF to which this interface belongs to. If the
interface is not in a VRF then it is 'Global'";
}
leaf ipv4 {
type inet:ip-address;
description
"IPv4 address configured on interface";
}
leaf ipv4-subnet-mask {
type inet:ip-address;
description
"IPv4 Subnet Mask";
}
leaf description {
type string;
description
"Interface description";
}
leaf mtu {
type uint32;
description
"Maximum transmission unit";
}
leaf input-security-acl {
type string;
description
"Input Security ACL";
}
leaf output-security-acl {
type string;
description
"Output Security ACL";
}
container v4-protocol-stats {
description
"IPv4 traffic statistics for this interface";
uses interfaces-ios-xe-oper:protocol-statistics;
}
container v6-protocol-stats {
description
"IPv6 traffic statistics for this interface";
uses interfaces-ios-xe-oper:protocol-statistics;
}
leaf bia-address {
type yang:mac-address;
description
"The burnt-in mac address that was associated with this
interface from manufacturing. This is only relevant for
interfaces that have the concept of burnt in ethernet
addresses, otherwise it is zero.";
}
leaf-list ipv6-addrs {
type inet:ip-address;
ordered-by user;
description
"A list of the IPv6 addresses associated with the interface.
This contains all the IPv6 addresses, including the link local
addresses, assigned to the interface";
}
list lag-aggregate-state {
key "aggregate-id";
description
"Operational state variables for logical
aggregate / LAG interfaces";
uses interfaces-ios-xe-oper:lag-aggregate-state;
}
leaf ipv4-tcp-adjust-mss {
type uint16;
description
"When ip tcp adjust-mss is configured, this value shows
the tcp mss, or the value is zero.";
}
leaf ipv6-tcp-adjust-mss {
type uint16;
description
"When ipv6 tcp adjust-mss is configured, this value shows
the tcp mss, or the value is zero.";
}
choice interface-class-choice {
description
"The broad class that the interface belongs";
case interface-class-ethernet {
container ether-state {
description
"The Ethernet state information";
uses interfaces-ios-xe-oper:ethernet-state;
}
container ether-stats {
description
"The Ethernet statistics";
uses interfaces-ios-xe-oper:ethernet-statistics;
}
}
case interface-class-t1e1serial {
container serial-state {
description
"The T1E1 serial state information";
uses interfaces-ios-xe-oper:t1e1-serial-state;
}
container serial-stats {
description
"The T1E1 statistics";
uses interfaces-ios-xe-oper:t1e1-serial-stats;
}
}
case interface-class-unspecified {
leaf intf-class-unspecified {
type boolean;
description
"No specific interface class information";
}
}
case interface-class-syncserial {
container syncserial-state {
description
"The synchronous serial state information";
uses interfaces-ios-xe-oper:sync-serial-state;
}
}
}
}
container interfaces {
config false;
description
"Operational state of interfaces";
list interface {
key "name";
description
"List of interfaces";
uses interfaces-ios-xe-oper:interface-state;
}
}
}
###Markdown
That's not so readable. Let's use a utility called ```pyang``` to get something a bit more readable.
###Code
from subprocess import Popen, PIPE, STDOUT
SCHEMA_TO_GET = 'Cisco-IOS-XE-interfaces-oper'
c = m.get_schema(SCHEMA_TO_GET)
# Simple pyang tree display
# p = Popen(['pyang', '-f', 'tree'], stdout=PIPE, stdin=PIPE, stderr=PIPE)
# Restrict display depth
# p = Popen(['pyang', '-f', 'tree', '--tree-depth', '2'], stdout=PIPE, stdin=PIPE, stderr=PIPE)
# Restrict display path
p = Popen(['pyang',
'-f', 'tree',
'--tree-path', 'interfaces/interface/statistics'],
stdout=PIPE, stdin=PIPE, stderr=PIPE)
# Push the data from the get_schema through pyang
stdout_data = p.communicate(input=c.data.encode())[0]
print(stdout_data.decode())
###Output
_____no_output_____
###Markdown
What About Config?The ncclient library provides for some simple operations. Let's skip thinking about schemas and stuff like that. Instead let's focus on config and getting end setting it. For that, ncclient provides two methods:* get_config - takes a target data store and an optional filter* edit_config - takes a target data store and an XML document with the edit request Getting ConfigLet's look at some simple requests...
###Code
c = m.get_config(source='running')
pretty_print(c)
###Output
_____no_output_____
###Markdown
Now let's add in a simple filter:
###Code
filter = '''
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<username/>
</native>
'''
with m.locked(target='running'):
c = m.get_config(source='running', filter=('subtree', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<username>
<name>vagrant</name>
<privilege>15</privilege>
<password>
<encryption>0</encryption>
<password>vagrant</password>
</password>
</username>
</native>
</data>
###Markdown
Retrieve Interface Config Data (Native Model)
###Code
filter = '''
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<interface/>
</native>
'''
c = m.get_config(source='running', filter=('subtree', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<interface>
<GigabitEthernet>
<name>1</name>
<ip>
<address>
<dhcp/>
</address>
<igmp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-igmp">
<explicit-tracking>false</explicit-tracking>
<proxy-service>false</proxy-service>
<unidirectional-link>false</unidirectional-link>
<v3lite>false</v3lite>
</igmp>
</ip>
<mop>
<enabled>false</enabled>
<sysid>false</sysid>
</mop>
<speed xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-ethernet">
<value-1000/>
</speed>
<negotiation xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-ethernet">
<auto>false</auto>
</negotiation>
</GigabitEthernet>
<GigabitEthernet>
<name>2</name>
<shutdown/>
<ip>
<igmp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-igmp">
<explicit-tracking>false</explicit-tracking>
<proxy-service>false</proxy-service>
<unidirectional-link>false</unidirectional-link>
<v3lite>false</v3lite>
</igmp>
</ip>
<mop>
<enabled>false</enabled>
<sysid>false</sysid>
</mop>
<negotiation xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-ethernet">
<auto>true</auto>
</negotiation>
</GigabitEthernet>
<GigabitEthernet>
<name>3</name>
<shutdown/>
<ip>
<igmp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-igmp">
<explicit-tracking>false</explicit-tracking>
<proxy-service>false</proxy-service>
<unidirectional-link>false</unidirectional-link>
<v3lite>false</v3lite>
</igmp>
</ip>
<mop>
<enabled>false</enabled>
<sysid>false</sysid>
</mop>
<negotiation xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-ethernet">
<auto>true</auto>
</negotiation>
</GigabitEthernet>
<GigabitEthernet>
<name>4</name>
<shutdown/>
<ip>
<igmp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-igmp">
<explicit-tracking>false</explicit-tracking>
<proxy-service>false</proxy-service>
<unidirectional-link>false</unidirectional-link>
<v3lite>false</v3lite>
</igmp>
</ip>
<mop>
<enabled>false</enabled>
<sysid>false</sysid>
</mop>
<negotiation xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-ethernet">
<auto>true</auto>
</negotiation>
</GigabitEthernet>
<Loopback>
<name>100</name>
<description>CLEU Demo</description>
<ip>
<address>
<primary>
<address>172.16.1.1</address>
<mask>255.255.255.255</mask>
</primary>
</address>
<igmp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-igmp">
<explicit-tracking>false</explicit-tracking>
<proxy-service>false</proxy-service>
<unidirectional-link>false</unidirectional-link>
<v3lite>false</v3lite>
</igmp>
</ip>
</Loopback>
</interface>
</native>
</data>
###Markdown
Retrieve Interface Data (Native Model) With XPath QueryAs well as subtree filters, **IOS-XE** supports XPath-based filters.
###Code
filter = '/native/interface/GigabitEthernet/name'
c = m.get_config(source='running', filter=('xpath', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<interface>
<GigabitEthernet>
<name>1</name>
</GigabitEthernet>
<GigabitEthernet>
<name>2</name>
</GigabitEthernet>
<GigabitEthernet>
<name>3</name>
</GigabitEthernet>
<GigabitEthernet>
<name>4</name>
</GigabitEthernet>
</interface>
</native>
</data>
###Markdown
Retrieve All BGP DataNow let's look at the BGP native model:
###Code
filter = '''
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp"/>
</router>
</native>
'''
c = m.get_config(source='running', filter=('subtree', filter))
pretty_print(c)
###Output
_____no_output_____
###Markdown
Look At A Specific BGP NeighborAnd can we look at a specific neighbor only? Say the one with id (address) ```2.2.2.3```?
###Code
filter = '''
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp">
<id>100</id>
<neighbor>
<id>2.2.2.3</id>
</neighbor>
<address-family>
<no-vrf>
<ipv4>
<af-name>unicast</af-name>
<ipv4-unicast>
<neighbor>
<id>2.2.2.3</id>
</neighbor>
</ipv4-unicast>
</ipv4>
</no-vrf>
</address-family>
</bgp>
</router>
</native>
'''
c = m.get_config(source='running', filter=('subtree', filter))
pretty_print(c)
###Output
_____no_output_____
###Markdown
Create New BGP NeighborOk, so, yes we can get a specific neighbor. Now, can we create a new neighbor? Let's create one with an id of `2.2.2.4`, with a remote-as of 666.
###Code
from ncclient.operations import TimeoutExpiredError
edit_data = '''
<config xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp">
<id>100</id>
<neighbor>
<id>2.2.2.4</id>
<remote-as>666</remote-as>
</neighbor>
<address-family>
<no-vrf>
<ipv4>
<af-name>unicast</af-name>
<ipv4-unicast>
<neighbor>
<id>2.2.2.4</id>
<activate/>
</neighbor>
</ipv4-unicast>
</ipv4>
</no-vrf>
</address-family>
</bgp>
</router>
</native>
</config>
'''
try:
with m.locked(target='running'):
edit_reply = m.edit_config(edit_data, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
_____no_output_____
###Markdown
Now let's pull back some of the data for that neighbor:
###Code
filter = '''
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp">
<id>100</id>
<neighbor>
<id>2.2.2.4</id>
</neighbor>
</bgp>
</router>
</native>
'''
c = m.get_config(source='running', filter=('subtree', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp">
<id xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">100</id>
<neighbor>
<id xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">2.2.2.4</id>
<remote-as>666</remote-as>
<timers>
<keepalive-interval>60</keepalive-interval>
<holdtime>180</holdtime>
</timers>
</neighbor>
</bgp>
</router>
</native>
</data>
###Markdown
Modify The BGP Neighbor DescriptionCan modify something in the neighbor we just created? Let's keep it simple and modify the description:
###Code
from ncclient.operations import TimeoutExpiredError
edit_data = '''
<config xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp">
<id>100</id>
<neighbor>
<id>2.2.2.4</id>
<description nc:operation="merge">*** MODIFIED DESCRIPTION ***</description>
</neighbor>
</bgp>
</router>
</native>
</config>
'''
try:
with m.locked(target='running'):
edit_reply = m.edit_config(edit_data, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
_____no_output_____
###Markdown
Delete A BGP Neighbor
###Code
from ncclient.operations import TimeoutExpiredError
from lxml.etree import XMLSyntaxError
edit_data = '''
<config xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<router>
<bgp xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-bgp">
<id>100</id>
<neighbor nc:operation="delete">
<id>2.2.2.4</id>
</neighbor>
<address-family>
<no-vrf>
<ipv4>
<af-name>unicast</af-name>
<ipv4-unicast>
<neighbor nc:operation="delete">
<id>2.2.2.4</id>
</neighbor>
</ipv4-unicast>
</ipv4>
</no-vrf>
</address-family>
</bgp>
</router>
</native>
</config>
'''
try:
with m.locked(target='running'):
edit_reply = m.edit_config(edit_data, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except XMLSyntaxError as e:
print(e)
print(e.args)
print(dir(e))
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
severity=error, tag=data-missing
{'type': 'application', 'tag': 'data-missing', 'severity': 'error', 'info': '<?xml version="1.0" encoding="UTF-8"?><error-info xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0"><bad-element>neighbor</bad-element>\n</error-info>\n', 'path': "\n /nc:rpc/nc:edit-config/nc:config/ios:native/ios:router/ios-bgp:bgp[ios-bgp:id='100']/ios-bgp:neighbor[ios-bgp:id='2.2.2.4']\n ", 'message': None}
###Markdown
Other Stuff Get interface operational data from native model:
###Code
filter = '''
<interfaces xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper">
<interface/>
</interfaces>
'''
c = m.get(filter=('subtree', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<interfaces xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper">
<interface>
<name>Control Plane</name>
<interface-type>iana-iftype-other</interface-type>
<admin-status>if-state-up</admin-status>
<oper-status>if-oper-state-ready</oper-status>
<last-change>2020-01-19T15:11:26.873+00:00</last-change>
<if-index>0</if-index>
<phys-address>00:00:00:00:00:00</phys-address>
<speed>10240000000</speed>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>0</in-octets>
<in-unicast-pkts>0</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>0</out-octets>
<out-unicast-pkts>0</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>0</out-octets-64>
</statistics>
<vrf/>
<description/>
<mtu>0</mtu>
<input-security-acl/>
<output-security-acl/>
<v4-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v4-protocol-stats>
<v6-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v6-protocol-stats>
<bia-address>00:00:00:00:00:00</bia-address>
<ipv4-tcp-adjust-mss>0</ipv4-tcp-adjust-mss>
<ipv6-tcp-adjust-mss>0</ipv6-tcp-adjust-mss>
<intf-class-unspecified>true</intf-class-unspecified>
</interface>
<interface>
<name>GigabitEthernet1</name>
<interface-type>iana-iftype-ethernet-csmacd</interface-type>
<admin-status>if-state-up</admin-status>
<oper-status>if-oper-state-ready</oper-status>
<last-change>2020-01-19T21:34:03.078+00:00</last-change>
<if-index>1</if-index>
<phys-address>08:00:27:2f:96:1b</phys-address>
<speed>1024000000</speed>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>795763</in-octets>
<in-unicast-pkts>7629</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>6678326</out-octets>
<out-unicast-pkts>6714</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>6678326</out-octets-64>
</statistics>
<vrf/>
<ipv4>10.0.2.15</ipv4>
<ipv4-subnet-mask>255.255.255.0</ipv4-subnet-mask>
<description/>
<mtu>1500</mtu>
<input-security-acl/>
<output-security-acl/>
<v4-protocol-stats>
<in-pkts>1725</in-pkts>
<in-octets>182181</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>1533</out-pkts>
<out-octets>1593544</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>1533</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v4-protocol-stats>
<v6-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v6-protocol-stats>
<bia-address>08:00:27:2f:96:1b</bia-address>
<ipv4-tcp-adjust-mss>0</ipv4-tcp-adjust-mss>
<ipv6-tcp-adjust-mss>0</ipv6-tcp-adjust-mss>
<ether-state>
<negotiated-duplex-mode>full-duplex</negotiated-duplex-mode>
<negotiated-port-speed>speed-1gb</negotiated-port-speed>
<auto-negotiate>false</auto-negotiate>
<enable-flow-control>false</enable-flow-control>
<media-type>ether-media-type-virtual</media-type>
</ether-state>
<ether-stats>
<in-mac-control-frames>0</in-mac-control-frames>
<in-mac-pause-frames>0</in-mac-pause-frames>
<in-oversize-frames>0</in-oversize-frames>
<in-jabber-frames>0</in-jabber-frames>
<in-fragment-frames>0</in-fragment-frames>
<in-8021q-frames>0</in-8021q-frames>
<out-mac-control-frames>0</out-mac-control-frames>
<out-mac-pause-frames>0</out-mac-pause-frames>
<out-8021q-frames>0</out-8021q-frames>
</ether-stats>
</interface>
<interface>
<name>GigabitEthernet2</name>
<interface-type>iana-iftype-ethernet-csmacd</interface-type>
<admin-status>if-state-down</admin-status>
<oper-status>if-oper-state-no-pass</oper-status>
<last-change>2020-01-19T15:11:28.941+00:00</last-change>
<if-index>2</if-index>
<phys-address>08:00:27:8f:64:dc</phys-address>
<speed>1024000000</speed>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>0</in-octets>
<in-unicast-pkts>0</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>0</out-octets>
<out-unicast-pkts>0</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>0</out-octets-64>
</statistics>
<vrf/>
<description/>
<mtu>1500</mtu>
<input-security-acl/>
<output-security-acl/>
<v4-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v4-protocol-stats>
<v6-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v6-protocol-stats>
<bia-address>08:00:27:8f:64:dc</bia-address>
<ipv4-tcp-adjust-mss>0</ipv4-tcp-adjust-mss>
<ipv6-tcp-adjust-mss>0</ipv6-tcp-adjust-mss>
<ether-state>
<negotiated-duplex-mode>full-duplex</negotiated-duplex-mode>
<negotiated-port-speed>speed-1gb</negotiated-port-speed>
<auto-negotiate>true</auto-negotiate>
<enable-flow-control>false</enable-flow-control>
<media-type>ether-media-type-virtual</media-type>
</ether-state>
<ether-stats>
<in-mac-control-frames>0</in-mac-control-frames>
<in-mac-pause-frames>0</in-mac-pause-frames>
<in-oversize-frames>0</in-oversize-frames>
<in-jabber-frames>0</in-jabber-frames>
<in-fragment-frames>0</in-fragment-frames>
<in-8021q-frames>0</in-8021q-frames>
<out-mac-control-frames>0</out-mac-control-frames>
<out-mac-pause-frames>0</out-mac-pause-frames>
<out-8021q-frames>0</out-8021q-frames>
</ether-stats>
</interface>
<interface>
<name>GigabitEthernet3</name>
<interface-type>iana-iftype-ethernet-csmacd</interface-type>
<admin-status>if-state-down</admin-status>
<oper-status>if-oper-state-no-pass</oper-status>
<last-change>2020-01-19T15:11:28.942+00:00</last-change>
<if-index>3</if-index>
<phys-address>08:00:27:72:d6:8f</phys-address>
<speed>1024000000</speed>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>0</in-octets>
<in-unicast-pkts>0</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>0</out-octets>
<out-unicast-pkts>0</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>0</out-octets-64>
</statistics>
<vrf/>
<description/>
<mtu>1500</mtu>
<input-security-acl/>
<output-security-acl/>
<v4-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v4-protocol-stats>
<v6-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v6-protocol-stats>
<bia-address>08:00:27:72:d6:8f</bia-address>
<ipv4-tcp-adjust-mss>0</ipv4-tcp-adjust-mss>
<ipv6-tcp-adjust-mss>0</ipv6-tcp-adjust-mss>
<ether-state>
<negotiated-duplex-mode>full-duplex</negotiated-duplex-mode>
<negotiated-port-speed>speed-1gb</negotiated-port-speed>
<auto-negotiate>true</auto-negotiate>
<enable-flow-control>false</enable-flow-control>
<media-type>ether-media-type-virtual</media-type>
</ether-state>
<ether-stats>
<in-mac-control-frames>0</in-mac-control-frames>
<in-mac-pause-frames>0</in-mac-pause-frames>
<in-oversize-frames>0</in-oversize-frames>
<in-jabber-frames>0</in-jabber-frames>
<in-fragment-frames>0</in-fragment-frames>
<in-8021q-frames>0</in-8021q-frames>
<out-mac-control-frames>0</out-mac-control-frames>
<out-mac-pause-frames>0</out-mac-pause-frames>
<out-8021q-frames>0</out-8021q-frames>
</ether-stats>
</interface>
<interface>
<name>GigabitEthernet4</name>
<interface-type>iana-iftype-ethernet-csmacd</interface-type>
<admin-status>if-state-down</admin-status>
<oper-status>if-oper-state-no-pass</oper-status>
<last-change>2020-01-19T15:11:28.943+00:00</last-change>
<if-index>4</if-index>
<phys-address>08:00:27:86:41:2a</phys-address>
<speed>1024000000</speed>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>0</in-octets>
<in-unicast-pkts>0</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>0</out-octets>
<out-unicast-pkts>0</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>0</out-octets-64>
</statistics>
<vrf/>
<description/>
<mtu>1500</mtu>
<input-security-acl/>
<output-security-acl/>
<v4-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v4-protocol-stats>
<v6-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v6-protocol-stats>
<bia-address>08:00:27:86:41:2a</bia-address>
<ipv4-tcp-adjust-mss>0</ipv4-tcp-adjust-mss>
<ipv6-tcp-adjust-mss>0</ipv6-tcp-adjust-mss>
<ether-state>
<negotiated-duplex-mode>full-duplex</negotiated-duplex-mode>
<negotiated-port-speed>speed-1gb</negotiated-port-speed>
<auto-negotiate>true</auto-negotiate>
<enable-flow-control>false</enable-flow-control>
<media-type>ether-media-type-virtual</media-type>
</ether-state>
<ether-stats>
<in-mac-control-frames>0</in-mac-control-frames>
<in-mac-pause-frames>0</in-mac-pause-frames>
<in-oversize-frames>0</in-oversize-frames>
<in-jabber-frames>0</in-jabber-frames>
<in-fragment-frames>0</in-fragment-frames>
<in-8021q-frames>0</in-8021q-frames>
<out-mac-control-frames>0</out-mac-control-frames>
<out-mac-pause-frames>0</out-mac-pause-frames>
<out-8021q-frames>0</out-8021q-frames>
</ether-stats>
</interface>
<interface>
<name>Loopback100</name>
<interface-type>iana-iftype-sw-loopback</interface-type>
<admin-status>if-state-up</admin-status>
<oper-status>if-oper-state-ready</oper-status>
<last-change>2020-01-19T21:49:44.163+00:00</last-change>
<if-index>7</if-index>
<phys-address>00:1e:49:ea:3c:00</phys-address>
<speed>8192000000</speed>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>0</in-octets>
<in-unicast-pkts>0</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>0</out-octets>
<out-unicast-pkts>0</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>0</out-octets-64>
</statistics>
<vrf/>
<ipv4>172.16.1.1</ipv4>
<ipv4-subnet-mask>255.255.255.255</ipv4-subnet-mask>
<description>CLEU Demo</description>
<mtu>1514</mtu>
<input-security-acl/>
<output-security-acl/>
<v4-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v4-protocol-stats>
<v6-protocol-stats>
<in-pkts>0</in-pkts>
<in-octets>0</in-octets>
<in-error-pkts>0</in-error-pkts>
<in-forwarded-pkts>0</in-forwarded-pkts>
<in-forwarded-octets>0</in-forwarded-octets>
<in-discarded-pkts>0</in-discarded-pkts>
<out-pkts>0</out-pkts>
<out-octets>0</out-octets>
<out-error-pkts>0</out-error-pkts>
<out-forwarded-pkts>0</out-forwarded-pkts>
<out-forwarded-octets>0</out-forwarded-octets>
<out-discarded-pkts>0</out-discarded-pkts>
</v6-protocol-stats>
<bia-address>00:00:00:00:00:00</bia-address>
<ipv4-tcp-adjust-mss>0</ipv4-tcp-adjust-mss>
<ipv6-tcp-adjust-mss>0</ipv6-tcp-adjust-mss>
<intf-class-unspecified>true</intf-class-unspecified>
</interface>
</interfaces>
</data>
###Markdown
Get all interface names
###Code
filter = '''
<interfaces xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper">
<interface>
<name/>
</interface>
</interfaces>
'''
c = m.get(filter=('subtree', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<interfaces xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper">
<interface>
<name>Control Plane</name>
</interface>
<interface>
<name>GigabitEthernet1</name>
</interface>
<interface>
<name>GigabitEthernet2</name>
</interface>
<interface>
<name>GigabitEthernet3</name>
</interface>
<interface>
<name>GigabitEthernet4</name>
</interface>
<interface>
<name>Loopback100</name>
</interface>
</interfaces>
</data>
###Markdown
Get the statistics from a specific interface:
###Code
filter = '''
<interfaces xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper">
<interface>
<name>GigabitEthernet1</name>
<statistics/>
</interface>
</interfaces>
'''
c = m.get(filter=('subtree', filter))
pretty_print(c)
###Output
<data xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<interfaces xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-interfaces-oper">
<interface>
<name xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">GigabitEthernet1</name>
<statistics>
<discontinuity-time>2020-01-19T15:09:37+00:00</discontinuity-time>
<in-octets>797761</in-octets>
<in-unicast-pkts>7650</in-unicast-pkts>
<in-broadcast-pkts>0</in-broadcast-pkts>
<in-multicast-pkts>0</in-multicast-pkts>
<in-discards>0</in-discards>
<in-errors>0</in-errors>
<in-unknown-protos>0</in-unknown-protos>
<out-octets>6697678</out-octets>
<out-unicast-pkts>6734</out-unicast-pkts>
<out-broadcast-pkts>0</out-broadcast-pkts>
<out-multicast-pkts>0</out-multicast-pkts>
<out-discards>0</out-discards>
<out-errors>0</out-errors>
<rx-pps>0</rx-pps>
<rx-kbps>0</rx-kbps>
<tx-pps>0</tx-pps>
<tx-kbps>0</tx-kbps>
<num-flaps>0</num-flaps>
<in-crc-errors>0</in-crc-errors>
<in-discards-64>0</in-discards-64>
<in-errors-64>0</in-errors-64>
<in-unknown-protos-64>0</in-unknown-protos-64>
<out-octets-64>6697678</out-octets-64>
</statistics>
</interface>
</interfaces>
</data>
###Markdown
Exercise Constraints Create a VRF and put an interface into it
###Code
from ncclient.operations import TimeoutExpiredError
edit_data = '''
<config xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<interface>
<GigabitEthernet>
<name>4</name>
<vrf>
<forwarding>TEST</forwarding>
</vrf>
<ip>
<address>
<primary>
<address>192.168.1.222</address>
<mask>255.255.255.0</mask>
</primary>
</address>
</ip>
</GigabitEthernet>
</interface>
<vrf>
<definition>
<name>TEST</name>
<address-family>
<ipv4/>
</address-family>
</definition>
</vrf>
</native>
</config>
'''
try:
with m.locked(target='running'):
edit_reply = m.edit_config(edit_data, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
_____no_output_____
###Markdown
Try to delete the VRF
###Code
from ncclient.operations import TimeoutExpiredError
edit_data = '''
<config xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<vrf>
<definition nc:operation="delete">
<name>TEST</name>
</definition>
</vrf>
</native>
</config>
'''
try:
with m.locked(target='running'):
edit_reply = m.edit_config(edit_data, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
_____no_output_____
###Markdown
Delete the VRF right
###Code
from ncclient.operations import TimeoutExpiredError
import textwrap
# get all interfaces where VRF "TEST" is bound
xpath = "/native/interface/*/vrf[forwarding='TEST']"
c = m.get(filter=('xpath', xpath))
# set operation=delete on every VRF found
for e in c.data.xpath('//*[local-name()="vrf"]'):
e.attrib['{urn:ietf:params:xml:ns:netconf:base:1.0}operation'] = "delete"
# find the node "native" and add it to a config node and then run an edit
config = etree.Element(
"config",
nsmap = {None: 'urn:ietf:params:xml:ns:netconf:base:1.0'})
config.append(c.data.xpath('//*[local-name()="native"]')[0])
# display what we will try to delete
print('What we will no try to delete:\n')
to_delete = etree.tostring(config, pretty_print=True).decode()
print(textwrap.indent(to_delete, ' ') )
# now delete the VRF from, the interface
try:
with m.locked(target='running'):
edit_reply = m.edit_config(config, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
_____no_output_____
###Markdown
Try put an interface into a non-existent VRF
###Code
from ncclient.operations import TimeoutExpiredError
edit_data = '''
<config xmlns="urn:ietf:params:xml:ns:netconf:base:1.0" xmlns:nc="urn:ietf:params:xml:ns:netconf:base:1.0">
<native xmlns="http://cisco.com/ns/yang/Cisco-IOS-XE-native">
<interface>
<GigabitEthernet>
<name>3</name>
<vrf>
<forwarding>DOES_NOT_EXIST</forwarding>
</vrf>
<ip>
<address>
<primary>
<address>192.168.2.222</address>
<mask>255.255.255.0</mask>
</primary>
</address>
</ip>
</GigabitEthernet>
</interface>
</native>
</config>
'''
try:
with m.locked(target='running'):
edit_reply = m.edit_config(edit_data, target='running', format='xml')
except TimeoutExpiredError as e:
print("Operation timeout!")
except Exception as e:
print("severity={}, tag={}".format(e.severity, e.tag))
print(e)
###Output
_____no_output_____
###Markdown
Alternate XML Creation
###Code
from lxml import etree
def create_interface_filter(intf_name):
interfaces = etree.Element(
"interfaces",
nsmap = {None: 'http://openconfig.net/yang/interfaces'})
interface = etree.SubElement(interfaces, "interface")
etree.SubElement(interface, 'name').text = intf_name
state = etree.SubElement(interface, 'state')
etree.SubElement(state, 'counters')
return interfaces
print(etree.tostring(create_interface_filter('MgmtEth0/RP0/CPU0/0'), pretty_print=True))
print(etree.tostring(create_interface_filter('GigabitEthernet1'), pretty_print=True))
###Output
_____no_output_____
###Markdown
Enable Debugging
###Code
import logging
handler = logging.StreamHandler()
for l in ['ncclient.transport.ssh', 'ncclient.transport.session', 'ncclient.operations.rpc']:
logger = logging.getLogger(l)
logger.addHandler(handler)
logger.setLevel(logging.DEBUG)
###Output
_____no_output_____ |
hello-jupyter/hello.ipynb | ###Markdown
Getting Started with PythonA [love letter](https://code.visualstudio.com/docs/python/python-tutorial) from Microsoft: “Let’s get started by creating the simplest "Hello World" Python application.”
###Code
msg = 'Hello World!'
print(msg)
###Output
Hello World!
###Markdown
The code above is taken directly from my hello [sample](../hello/hello.py).Of course, for the sake of completion, I would like to run the plottting [sample](../hello/standardplot.py) as well. However this will not work. It will produce unexpected output as described in a StackOverflow.com [post](https://stackoverflow.com/questions/19410042/how-to-make-ipython-notebook-matplotlib-plot-inline). The solution to this problem introduces the concept of the `inline` backend ([magic command](http://ipython.readthedocs.io/en/stable/interactive/magics.html?highlight=backendsmagic-matplotlib)):
###Code
%matplotlib inline
#%%
import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np
x = np.linspace(0, 20, 100)
plt.plot(x, np.sin(x))
plt.show()
###Output
_____no_output_____ |
header_footer/biosignalsnotebooks_environment/categories/Connect/pairing_device.ipynb | ###Markdown
<div id="image_img" class="header_image_14"> Pairing a Device at Windows 10 [biosignalsplux] Difficulty Level: Tags connect&9729;pairing&9729;biosignalsplux&9729;bluetooth&9729;windows10 For acquiring physiological data there are three systems in interaction: human body (system under exploration); signal acquisition system/sensors (system responsible for collecting data from the human body) and computer/mobile device (system that receives and processes the data collected by the acquisition system).Plux&39;s acquisition systems (biosignalsplux and bitalino ) valorize wireless communication between the acquisition system and computer (using Bluetooth technology).The current Jupyter Notebook is intended to explain how Plux&39;s acquisition systems (biosignalsplux in our example) can be quickly connect to a computer in order to ensure future real-time acquisitions through OpenSignals . 0 - Before starting an acquisition it is mandatory that your Plux acquisition system (in our case biosignalsplux) is paired with the computer. It will be described, in the following steps, relevant instructions to fulfill this prerequisite ! The example is focused on Microsoft Windows 10 Users. For Mac users there are some relevant pages explaining how a Bluetooth device, like biosignalsplux can be paired. https://support.apple.com/en-us/HT201171 https://support.apple.com/guide/mac-help/connect-a-bluetooth-device-blth1004/mac*It is possible to use the internal Bluetooth of your computer, however to improve the connectivity and have more freedom of movement and avoid communication losses, we strongly recommend to use a Bluetooth Dongle like the one available at Plux Store . 1 - Turn on your biosignalsplux device 2 - Navigate until the Bluetooth communication system windowThis action is achievable through the "Start Menu" icon [bottom-left corner of the screen ] >> Settings >> Devices >> Bluetooth & other devices 3 - Enable Bluetooth communication system in your computer 4 - Click on "Add Bluetooth or other device"We are reaching the pairing stage between our computer and biosignalsplux device 5 - Click on "Bluetooth" option and wait for the list of available devices 6 - Find and click on the entry labeled with "biosignalsplux" 7 - Type the predefined password >> "123" Mission almost accomplished ! Now you should have your biosignalsplux device ready to use in OpenSignals software. 8 - Execute OpenSignals application in order to get access to the main page 9 - Press the "Device Configuration" icon for enabling your device 10 - Check which is your device mac-address (at the back of your device) 11 - Find your device on the "Device Configuration" pageYou should press the "Search" icon in order to the previously added device (through Windows interface) can appear For finishing the pairing between biosignalsplux and the computer, it is only necessary to click on the device box. When a set of configurable options appear the pairing process is officially completed ! The previous instructions represents a reasonable interactive guide for establishing a connection between Plux&39;s devices and your computer.If you follow these steps, now you are connected with biosignalsplux, which can be the start of an amazing journey ! We hope that you have enjoyed this guide. biosignalsnotebooks is an environment in continuous expansion, so don't stop your journey and learn more with the remaining Notebooks ! &9740; Project Presentation &9740; GitHub Repository &9740; How to install biosignalsnotebooks Python package ? &9740; Signal Library &9740; Notebook Categories &9740; Notebooks by Difficulty &9740; Notebooks by Signal Type &9740; Notebooks by Tag
###Code
from biosignalsnotebooks.__notebook_support__ import css_style_apply
css_style_apply()
%%html
<script>
// AUTORUN ALL CELLS ON NOTEBOOK-LOAD!
require(
['base/js/namespace', 'jquery'],
function(jupyter, $) {
$(jupyter.events).on("kernel_ready.Kernel", function () {
console.log("Auto-running all cells-below...");
jupyter.actions.call('jupyter-notebook:run-all-cells-below');
jupyter.actions.call('jupyter-notebook:save-notebook');
});
}
);
</script>
###Output
_____no_output_____ |
immune_macrophages_GO.ipynb | ###Markdown
Functional characterization of novel macrophage phenotypes
###Code
import numpy as np
import pandas as pd
import scanpy as sc
import os
import sys
import warnings
import anndata
warnings.filterwarnings('ignore')
def MovePlots(plotpattern, subplotdir):
os.system('mkdir -p '+str(sc.settings.figdir)+'/'+subplotdir)
os.system('mv '+str(sc.settings.figdir)+'/*'+plotpattern+'** '+str(sc.settings.figdir)+'/'+subplotdir)
sc.settings.verbosity = 3 # verbosity: errors (0), warnings (1), info (2), hints (3)
sc.settings.figdir = './final-figures/merged/myeloid/functional_analysis/'
sc.logging.print_versions()
sc.settings.set_figure_params(dpi=80) # low dpi (dots per inch) yields small inline figures
sys.executable
###Output
WARNING: If you miss a compact list, please try `print_header`!
###Markdown
Took this function from: https://github.com/theislab/single-cell-tutorial/blob/master/latest_notebook/gprofiler_plotting.py
###Code
# Plotting functions - 'GProfiler-official version'
import matplotlib.pyplot as plt
import seaborn as sb
from matplotlib import colors
from matplotlib import rcParams
def scale_data_5_75(data):
mind = np.min(data)
maxd = np.max(data)
if maxd == mind:
maxd=maxd+1
mind=mind-1
drange = maxd - mind
return ((((data - mind)/drange*0.70)+0.05)*100)
def plot_enrich(data, n_terms=9, save=False):
# Test data input
if not isinstance(data, pd.DataFrame):
raise ValueError('Please input a Pandas Dataframe output by gprofiler.')
if not np.all([term in data.columns for term in ['p_value', 'name', 'intersection_size']]):
raise TypeError('The data frame {} does not contain enrichment results from gprofiler.'.format(data))
data_to_plot = data.iloc[:n_terms,:].copy()
data_to_plot['go.id'] = data_to_plot.index
min_pval = data_to_plot['p_value'].min()
max_pval = data_to_plot['p_value'].max()
# Scale intersection_size to be between 5 and 75 for plotting
#Note: this is done as calibration was done for values between 5 and 75
data_to_plot['scaled.overlap'] = scale_data_5_75(data_to_plot['intersection_size'])
norm = colors.LogNorm(min_pval, max_pval)
sm = plt.cm.ScalarMappable(cmap="OrRd", norm=norm)
sm.set_array([])
rcParams.update({'font.size': 12, 'font.weight': 'normal'})
sb.set(style="whitegrid")
path = plt.scatter(x='recall', y="name", c='p_value', cmap='OrRd',
norm=colors.LogNorm(min_pval, max_pval),
data=data_to_plot, linewidth=1, edgecolor="grey",
s=[(i+10)**1.5 for i in data_to_plot['scaled.overlap']])
ax = plt.gca()
ax.invert_yaxis()
ax.set_ylabel('')
ax.set_xlabel('Gene ratio', fontsize=12, fontweight='normal')
ax.xaxis.grid(False)
ax.yaxis.grid(True)
# Shrink current axis by 20%
box = ax.get_position()
ax.set_position([box.x0, box.y0, box.width * 0.8, box.height])
# Get tick marks for this plot
#Note: 6 ticks maximum
min_tick = np.floor(np.log10(min_pval)).astype(int)
max_tick = np.ceil(np.log10(max_pval)).astype(int)
tick_step = np.ceil((max_tick - min_tick)/6).astype(int)
# Ensure no 0 values
if tick_step == 0:
tick_step = 1
min_tick = max_tick-1
ticks_vals = [10**i for i in range(max_tick, min_tick-1, -tick_step)]
ticks_labs = ['$10^{'+str(i)+'}$' for i in range(max_tick, min_tick-1, -tick_step)]
#Colorbar
fig = plt.gcf()
cbaxes = fig.add_axes([0.8, 0.10, 0.03, 0.4])
cbar = ax.figure.colorbar(sm, ticks=ticks_vals, shrink=0.2, anchor=(0,0.1), cax=cbaxes)
cbar.ax.set_yticklabels(ticks_labs)
cbar.set_label("Adjusted p-value", fontsize=12, fontweight='normal')
#Size legend
min_olap = data_to_plot['intersection_size'].min()
max_olap = data_to_plot['intersection_size'].max()
olap_range = max_olap - min_olap
#Note: approximate scaled 5, 25, 50, 75 values are calculated
# and then rounded to nearest number divisible by 5
size_leg_vals = [np.round(i/5)*5 for i in
[min_olap, min_olap+(20/70)*olap_range, min_olap+(45/70)*olap_range, max_olap]]
size_leg_scaled_vals = scale_data_5_75(size_leg_vals)
l1 = plt.scatter([],[], s=(size_leg_scaled_vals[0]+10)**1.5, edgecolors='none', color='black')
l2 = plt.scatter([],[], s=(size_leg_scaled_vals[1]+10)**1.5, edgecolors='none', color='black')
l3 = plt.scatter([],[], s=(size_leg_scaled_vals[2]+10)**1.5, edgecolors='none', color='black')
l4 = plt.scatter([],[], s=(size_leg_scaled_vals[3]+10)**1.5, edgecolors='none', color='black')
labels = [str(int(i)) for i in size_leg_vals]
leg = plt.legend([l1, l2, l3, l4], labels, ncol=1, frameon=False, fontsize=12,
handlelength=1, loc = 'center left', borderpad = 1, labelspacing = 1.4,
handletextpad=2, title='Gene overlap', scatterpoints = 1, bbox_to_anchor=(-2, 1.5),
facecolor='black')
if save:
plt.savefig(save, dpi=300, format='pdf')
plt.show()
from gprofiler import GProfiler
###Output
_____no_output_____
###Markdown
Load data
###Code
path_to_gonads = '/nfs/team292/vl6/immune_fetal_gonads/'
adata = sc.read(path_to_gonads + 'macrophages.h5ad')
sc.pl.umap(adata, color = 'clusters')
adata.obs.to_csv(path_to_gonads + 'macrophage_annotations.csv')
###Output
_____no_output_____
###Markdown
Differential expression analysis
###Code
import rpy2.rinterface_lib.callbacks
import logging
# Ignore R warning messages
#Note: this can be commented out to get more verbose R output
rpy2.rinterface_lib.callbacks.logger.setLevel(logging.ERROR)
import anndata2ri
anndata2ri.activate()
%load_ext rpy2.ipython
import anndata
adata = anndata.AnnData(X = adata.raw.X, var = adata.raw.var, obs = adata.obs)
%%R -i adata
adata
%%R -o mrks
library(SoupX)
counts <- assay(adata, "X")
colnames(counts) <- colnames(adata)
rownames(counts) <- rownames(adata)
mrks = quickMarkers(counts, colData(adata)$clusters, N = 100) # where clusters is the cell type assignment and 20 means I want the top 20 genes per cluster that pass the hypergeometric test
mrks
mrks.to_csv("/home/jovyan/Macrophages_markersTFIDF.csv")
tissue_repair_markers = mrks[mrks['cluster'] == 'tissue-repair mac']['gene'].to_list()
ifng_induced_markers = mrks[mrks['cluster'] == 'ifn-induced mac']['gene'].to_list()
microglialike_markers = mrks[mrks['cluster'] == 'microglia-like mac']['gene'].to_list()
osteoclastlike_markers = mrks[mrks['cluster'] == 'SIGLEC15+ mac']['gene'].to_list()
len(tissue_repair_markers)
###Output
_____no_output_____
###Markdown
gProfiler (GO-Biological Process enrichment)1. Tissue-repair mac
###Code
gp = GProfiler(
user_agent='ExampleTool', #optional user agent
return_dataframe=True, #return pandas dataframe or plain python structures
)
tissue_repair_enrichment = gp.profile(organism='hsapiens', sources=['GO:BP'], user_threshold=0.1,
significance_threshold_method='fdr',
background=adata.var_names.tolist(),
query=tissue_repair_markers)
tissue_repair_results = tissue_repair_enrichment.set_index('native').sort_values('p_value').iloc[:,[2,5,7,10,1]]
pd.set_option("display.max_colwidth", 800)
tissue_repair_results.iloc[:30,:]
selected_terms_tissue_repair = ['response to stimulus', 'inflammatory response',
'response to oxygen-containing compound']
selected_terms_tissue_repair = tissue_repair_results[tissue_repair_results['name'].isin(selected_terms_tissue_repair)]
selected_terms_tissue_repair
selected_terms_tissue_repair.reset_index(level=0, inplace=True)
selected_terms_tissue_repair.style.hide_index()
style = selected_terms_tissue_repair.style.set_table_styles([{'selector' : '',
'props' : [('border',
'3px solid #d9abb7'), ('background-color', '#d9abb7')]}])
style.hide_index()
tissue_repair_results.iloc[:100,:].to_csv('/home/jovyan/TissueRepair_GO.csv')
###Output
_____no_output_____
###Markdown
2. IFN-induced mac
###Code
gp = GProfiler(
user_agent='ExampleTool', #optional user agent
return_dataframe=True, #return pandas dataframe or plain python structures
)
ifn_induced_enrichment = gp.profile(organism='hsapiens', sources=['GO:BP'], user_threshold=0.1,
significance_threshold_method='fdr',
background=adata.var_names.tolist(),
query=ifng_induced_markers)
ifn_induced_results = ifn_induced_enrichment.set_index('native').sort_values('p_value').iloc[:,[2,5,7,10,1]]
pd.set_option("display.max_colwidth", 800)
ifn_induced_results.iloc[:30,:]
selected_terms_IFN = ['response to cytokine', 'type I interferon signaling pathway',
'response to interferon-gamma']
selected_results_IFN = ifn_induced_results[ifn_induced_results['name'].isin(selected_terms_IFN)]
selected_results_IFN
selected_results_IFN.reset_index(level=0, inplace=True)
selected_results_IFN
selected_results_IFN.style.hide_index()
style = selected_results_IFN.style.set_table_styles([{'selector' : '',
'props' : [('border',
'3px solid #e64e74'), ('background-color', '#e64e74')]}])
style.hide_index()
ifn_induced_results.iloc[:100,:].to_csv('/home/jovyan/IFNInduced_GO.csv')
###Output
_____no_output_____
###Markdown
3. Microglia-like mac
###Code
gp = GProfiler(
user_agent='ExampleTool', #optional user agent
return_dataframe=True, #return pandas dataframe or plain python structures
)
microglia_enrichment = gp.profile(organism='hsapiens', sources=['GO:BP'], user_threshold=0.1,
significance_threshold_method='fdr',
background=adata.var_names.tolist(),
query=microglialike_markers)
microglia_results = microglia_enrichment.set_index('native').sort_values('p_value').iloc[:,[2,5,7,10,1]]
pd.set_option("display.max_colwidth", 800)
microglia_results.iloc[:30,:]
selected_terms_microglia = ['regulation of microglial cell migration', 'regulation of glial cell migration',
'positive regulation of macrophage migration']
selected_results_microglia = microglia_results[microglia_results['name'].isin(selected_terms_microglia)]
selected_results_microglia
selected_results_microglia.reset_index(level=0, inplace=True)
style1 = selected_results_microglia.style.set_table_styles([{'selector' : '',
'props' : [('border',
'3px solid #d9a5c3'), ('background-color', '#d9a5c3')]}])
style1.hide_index()
microglia_results.iloc[:100,:].to_csv('/home/jovyan/Microglia_GO.csv')
###Output
_____no_output_____
###Markdown
4. Osteoclast-like mac
###Code
gp = GProfiler(
user_agent='ExampleTool', #optional user agent
return_dataframe=True, #return pandas dataframe or plain python structures
)
osteoclast_enrichment = gp.profile(organism='hsapiens', sources=['GO:BP'], user_threshold=0.1,
significance_threshold_method='fdr',
background=adata.var_names.tolist(),
query=osteoclastlike_markers)
osteoclast_results = osteoclast_enrichment.set_index('native').sort_values('p_value').iloc[:,[2,5,7,10,1]]
pd.set_option("display.max_colwidth", 800)
osteoclast_results.iloc[:30,:]
selected_terms_osteoclast = ['extracellular matrix organization', 'bone remodeling',
'osteoclast differentiation']
selected_results_osteoclast = osteoclast_results[osteoclast_results['name'].isin(selected_terms_osteoclast)]
selected_results_osteoclast
selected_results_osteoclast.reset_index(level=0, inplace=True)
style2 = selected_results_osteoclast.style.set_table_styles([{'selector' : '',
'props' : [('border',
'3px solid #cc8fdb'), ('background-color', '#cc8fdb')]}])
style2.hide_index()
osteoclast_results.iloc[:100,:].to_csv('/home/jovyan/SIGLEC15_GO.csv')
###Output
_____no_output_____ |
utf-8''module_3_assignment.ipynb | ###Markdown
Module 3 AssignmentYour objective in this assignment is to implement a tennis ball detector using a pre-trained image classification network from GluonCV. We'll step through the pipeline, from loading and transforming an input image, to loading and using a pre-trained model. Since we're only interested in detecting tennis balls, this is a binary classification problem, which is slightly different to the multi-class classification setup we've seen so far. 0) SetupWe start with some initial setup: importing packages and setting the path to the data.
###Code
import mxnet as mx
import gluoncv as gcv
import matplotlib.pyplot as plt
import numpy as np
import os
from pathlib import Path
M3_DATA = Path(os.getenv('DATA_DIR', '../../data'), 'module_3')
M3_IMAGES = Path(M3_DATA, 'images')
M3_MODELS = Path(M3_DATA, 'models')
###Output
_____no_output_____
###Markdown
1) Loading an imageYour first task is to implement a function that loads an image from disk given a filepath.It should return an 8-bit image array, that's in MXNet's NDArray format and in HWC layout (i.e. height, width then channel).
###Code
def load_image(filepath):
"""
Should load image from disk.
:param filepath: relative or absolute filepath to RGB image file in JPG format.
:type filepath: str
:return: an array with pixel intensities (in HWC layout).
:rtype: mx.nd.NDArray
"""
# YOUR CODE HERE
image = mx.image.imread(filepath)
return image
test_filepath = Path(M3_IMAGES, 'ben-hershey-VEW78A1YZ6I-unsplash.jpg')
test_output = load_image(test_filepath)
assert test_output.shape[2] == 3 # RGB
assert test_output.dtype == np.uint8 # 0 - 255
assert isinstance(test_output, mx.nd.NDArray) # MXNet NDArray, not NumPy Array.
test_output.shape
###Output
_____no_output_____
###Markdown
2) Transforming an imageUp next, you should transform the image so it can be used as input to the pre-trained network.Since we're going to use an ImageNet pre-trained network, we need to follow the same steps used for ImageNet pre-training.See the docstring for more details, but don't forget that GluonCV contains a number of utilities and helper functions to make your life easier! Check out the preset transforms.
###Code
def transform_image(array):
"""
Should transform image by:
1) Resizing the shortest dimension to 224. e.g (448, 1792) -> (224, 896).
2) Cropping to a center square of dimension (224, 224).
3) Converting the image from HWC layout to CHW layout.
4) Normalizing the image using ImageNet statistics (i.e. per colour channel mean and variance).
5) Creating a batch of 1 image.
:param filepath: array (in HWC layout).
:type filepath: mx.nd.NDArray
:return: a batch of a single transformed images (in NCHW layout)
:rtype: mx.nd.NDArray
"""
transformed_img = gcv.data.transforms.presets.imagenet.transform_eval(array)
# YOUR CODE HERE
return transformed_img
transformed_test_output = transform_image(test_output)
assert transformed_test_output.shape == (1, 3, 224, 224)
assert transformed_test_output.dtype == np.float32
transformed_test_output.shape
###Output
_____no_output_____
###Markdown
3) Loading a modelWith the image loaded and transformed, you now need to load a pre-trained classification model.Choose a MobileNet 1.0 image classification model that's been pre-trained on ImageNet.**CAUTION!**: Although the notebook interface has internet connectivity, the **autograders are not permitted to access the internet**. We have already downloaded the correct models and data for you to use so you don't need access to the internet. However, you do need to specify the correct path to the models when loading a model from the Gluon CV Model Zoo using `get_model` or otherwise. Set the `root` parameter to `M3_MODELS`. As an example, you should have something similar to `gcv.model_zoo.get_model(..., root=M3_MODELS)`. Usually, in the real world, you have internet access, so setting the `root` parameter isn't required (and it's set to `~/.mxnet` by default).
###Code
def load_pretrained_classification_network():
"""
Loads a MobileNet 1.0 network that's been pre-trained on ImageNet.
:return: a pre-trained network
:rtype: mx.gluon.Block
"""
# YOUR CODE HERE
model = gcv.model_zoo.get_model('MobileNet1.0', pretrained=True, root = M3_MODELS)
return model
network = load_pretrained_classification_network()
assert isinstance(network, mx.gluon.Block), 'Model should be a Gluon Block'
assert network.name.startswith('mobilenet'), 'Select MobileNet'
params = network.collect_params(select=network.name + '_conv0_weight')
assert list(params.items())[0][1].shape[0] == 32, 'Select MobileNet1.0'
#3network
###Output
_____no_output_____
###Markdown
4) Using a modelYour next task is to pass your transformed image through the network to obtain predicted probabilities for all ImageNet classes.We'll ignore the requirement of creating just a tennis ball classifier for now.**Hint 1**: Don't forget that you're typically working with a batch of images, even when you only have one image.**Hint 2**: Remember that the direct outputs of our network aren't probabilities.
###Code
def predict_probabilities(network, data):
"""
Should return the predicted probabilities of ImageNet classes for the given image.
:param network: pre-trained image classification model
:type network: mx.gluon.Block
:param data: batch of transformed images of shape (1, 3, 224, 224)
:type data: mx.nd.NDArray
:return: array of probabilities of shape (1000,)
:rtype: mx.nd.NDArray
"""
# YOUR CODE HERE
prediction = network(data)
prediction = prediction[0]
probability = mx.nd.softmax(prediction)
return probability
pred_probas = predict_probabilities(network, transformed_test_output)
assert pred_probas.shape == (1000,)
np.testing.assert_almost_equal(pred_probas.sum().asscalar(), 1, decimal=5)
assert pred_probas.dtype == np.float32
pred_probas
pred_probas.shape
pred_probas.dtype
###Output
_____no_output_____
###Markdown
5) Finding Class LabelSince we're only interested in tennis ball classification for now, we need a method of finding the probability associated with tennis ball out of the 1000 classes.You should implement a function that returns the index of a given class label (e.g. `admiral` is index `321`)**Hint**: you're allowed to use variables that are defined globally on this occasion. You should think about which objects that have been previously defined has a list of class labels.
###Code
def find_class_idx(label):
"""
Should return the class index of a particular label.
:param label: label of class
:type label: str
:return: class index
:rtype: int
"""
# YOUR CODE HERE
for i in range(len(network.classes)):
if label == network.classes[i]:
return i
assert find_class_idx('tennis ball') == 852
assert find_class_idx('spiny lobster') == 123
assert find_class_idx('admiral') == 321
###Output
_____no_output_____
###Markdown
6) Slice Tennis Ball ClassUsing the above function to find the correct index for tennis ball, you should implement a function to slice the calculated probability for tennis ball from the 1000 class probabilities calculated by the network. It should also convert the probability from MXNet `NDArray` to a NumPy `float32`.We'll use this for our confidence score that the image is a tennis ball.
###Code
def slice_tennis_ball_class(pred_probas):
"""
Extracts the probability associated with tennis ball.
:param pred_probas: array of ImageNet probabilities of shape (1000,)
:type pred_probas: mx.nd.NDArray
:return: probability of tennis ball
:rtype: np.float32
"""
# YOUR CODE HERE
tennis_prob = pred_probas[find_class_idx('tennis ball')]
return tennis_prob.astype('float32').asscalar()
pred_proba_tennis_ball = slice_tennis_ball_class(pred_probas)
assert isinstance(pred_proba_tennis_ball, np.float32)
np.testing.assert_almost_equal(pred_proba_tennis_ball, 0.9987876, decimal=3)
pred_proba_tennis_ball
###Output
_____no_output_____
###Markdown
7) Classify Tennis Ball ImagesWe'll finish this assignment by bringing all of the components together and creating a `TennisBallClassifier` to classify images. You should implement the entire classification pipeline inside the `classify` function using the functions defined earlier on in the assignment. You should notice that the pre-trained model is loaded once during initialization, and then it should be used inside the `classify` method.
###Code
class TennisBallClassifier():
def __init__(self):
self._network = load_pretrained_classification_network()
def classify(self, filepath):
# YOUR CODE HERE
image = load_image(filepath)
transformed_image = transform_image(image)
self._visualize(transformed_image)
# YOUR CODE HERE
pred_probas = predict_probabilities(self._network, transformed_image)
pred_proba = slice_tennis_ball_class(pred_probas)
print('{0:.2%} confidence that image is a tennis ball.'.format(pred_proba))
return pred_proba
def _visualize(self, transformed_image):
"""
Since the transformed_image is in NCHW layout and the values are normalized,
this method slices and transposes to give CHW as required by matplotlib,
and scales (-2, +2) to (0, 255) linearly.
"""
chw_image = transformed_image[0].transpose((1,2,0))
chw_image = ((chw_image * 64) + 128).clip(0, 255).astype('uint8')
plt.imshow(chw_image.asnumpy())
classifier = TennisBallClassifier()
filepath = Path(M3_IMAGES, 'erik-mclean-D23_XPbsx-8-unsplash.jpg')
pred_proba = classifier.classify(filepath)
np.testing.assert_almost_equal(pred_proba, 2.0355723e-05, decimal=3)
filepath = Path(M3_IMAGES, 'marvin-ronsdorf-CA998Anw2Lg-unsplash.jpg')
pred_proba = classifier.classify(filepath)
np.testing.assert_almost_equal(pred_proba, 0.9988895, decimal=3)
###Output
99.92% confidence that image is a tennis ball.
|
docs/pages/examples/minimum_energy_fast.ipynb | ###Markdown
Energy settings
###Code
# setup states
n_nodes = A.shape[0]
n_states = int(n_nodes/10)
state_size = int(n_nodes/n_states)
states = np.array([])
for i in np.arange(n_states):
states = np.append(states, np.ones(state_size) * i)
states = states.astype(int)
print(states)
###Output
[ 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 2 2 2 2
2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4 4
4 4 5 5 5 5 5 5 5 5 5 5 6 6 6 6 6 6 6 6 6 6 7 7
7 7 7 7 7 7 7 7 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9
9 9 9 9 10 10 10 10 10 10 10 10 10 10 11 11 11 11 11 11 11 11 11 11
12 12 12 12 12 12 12 12 12 12 13 13 13 13 13 13 13 13 13 13 14 14 14 14
14 14 14 14 14 14 15 15 15 15 15 15 15 15 15 15 16 16 16 16 16 16 16 16
16 16 17 17 17 17 17 17 17 17 17 17 18 18 18 18 18 18 18 18 18 18 19 19
19 19 19 19 19 19 19 19]
###Markdown
Minimum energy
###Code
from network_control.utils import matrix_normalization
from network_control.energies import minimum_energy
# settings
# time horizon
T = 1
# set all nodes as control nodes
B = np.eye(n_nodes)
# normalize A matrix for a continuous-time system
A = matrix_normalization(A, version='continuous')
import time
start_time = time.time() # start timer
e = np.zeros((n_states, n_states, n_nodes))
for i in np.arange(n_states):
x0 = states == i # get ith initial state
for j in np.arange(n_states):
xf = states == j # get jth target state
m_x, m_u, n_err = minimum_energy(A, T, B, x0, xf)
e[i, j, :] = np.sum(np.square(m_u), axis=0)
end_time = time.time() # stop timer
elapsed_time = end_time - start_time
print('time elapsed in seconds: {:.2f}'.format(elapsed_time)) # print elapsed time
###Output
time elapsed in seconds: 38.15
###Markdown
Minimum energy fast
###Code
from network_control.utils import expand_states
x0_mat, xf_mat = expand_states(states)
print(x0_mat.shape, xf_mat.shape)
from network_control.energies import minimum_energy_fast
start_time = time.time() # start timer
e_fast = minimum_energy_fast(A, T, B, x0_mat, xf_mat)
e_fast = e_fast.transpose().reshape(n_states, n_states, n_nodes)
end_time = time.time() # stop timer
elapsed_time = end_time - start_time
print('time elapsed in seconds: {:.2f}'.format(elapsed_time)) # print elapsed time
print(e.shape)
print(e_fast.shape)
###Output
(20, 20, 200)
(20, 20, 200)
###Markdown
Plots
###Code
import matplotlib.pyplot as plt
import seaborn as sns
from network_control.plotting import set_plotting_params, reg_plot
set_plotting_params()
# sum energy over regions
e_sum = np.sum(e, axis=2)
e_fast_sum = np.sum(e_fast, axis=2)
# compute correlations across regional energy for each transition separately
r = np.zeros((n_states, n_states))
for i in np.arange(n_states):
for j in np.arange(n_states):
r[i, j] = sp.stats.pearsonr(e[i, j, :], e_fast[i, j, :])[0]
# plot
f, ax = plt.subplots(1, 2, figsize=(5, 2.5))
# correlation between whole-brain energy across state transitions
mask = ~np.eye(n_states, dtype=bool)
indices = np.where(mask)
reg_plot(x=e_sum[indices], y=e_fast_sum[indices], xlabel='minumum energy', ylabel='minumum energy (fast)', ax=ax[0],
add_spearman=True, kdeplot=False, regplot=False)
# energy correlated across regions for each state transition separately
sns.heatmap(r, square=True, ax=ax[1], cbar_kws={"shrink": 0.80})
ax[1].set_ylabel("initial states", labelpad=-1)
ax[1].set_xlabel("target states", labelpad=-1)
ax[1].set_yticklabels('')
ax[1].set_xticklabels('')
ax[1].tick_params(pad=-2.5)
plt.show()
f.savefig('minimum_energy_fast', dpi=300, bbox_inches='tight', pad_inches=0.1)
plt.close()
###Output
_____no_output_____ |
Course 2- NLP with probabilistic Model/Labs/NLP Week 3 Lab/NLP_C2_W3_lecture_nb_03.ipynb | ###Markdown
Out of vocabulary words (OOV) VocabularyIn the video about the out of vocabulary words, you saw that the first step in dealing with the unknown words is to decide which words belong to the vocabulary. In the code assignment, you will try the method based on minimum frequency - all words appearing in the training set with frequency >= minimum frequency are added to the vocabulary.Here is a code for the other method, where the target size of the vocabulary is known in advance and the vocabulary is filled with words based on their frequency in the training set.
###Code
# build the vocabulary from M most frequent words
# use Counter object from the collections library to find M most common words
from collections import Counter
# the target size of the vocabulary
M = 3
# pre-calculated word counts
# Counter could be used to build this dictionary from the source corpus
word_counts = {'happy': 5, 'because': 3, 'i': 2, 'am': 2, 'learning': 3, '.': 1}
vocabulary = Counter(word_counts).most_common(M)
# remove the frequencies and leave just the words
vocabulary = [w[0] for w in vocabulary]
print(f"the new vocabulary containing {M} most frequent words: {vocabulary}\n")
###Output
_____no_output_____
###Markdown
Now that the vocabulary is ready, you can use it to replace the OOV words with $$ as you saw in the lecture.
###Code
# test if words in the input sentences are in the vocabulary, if OOV, print <UNK>
sentence = ['am', 'i', 'learning']
output_sentence = []
print(f"input sentence: {sentence}")
for w in sentence:
# test if word w is in vocabulary
if w in vocabulary:
output_sentence.append(w)
else:
output_sentence.append('<UNK>')
print(f"output sentence: {output_sentence}")
###Output
_____no_output_____
###Markdown
When building the vocabulary in the code assignment, you will need to know how to iterate through the word counts dictionary. Here is an example of a similar task showing how to go through all the word counts and print out only the words with the frequency equal to f.
###Code
# iterate through all word counts and print words with given frequency f
f = 3
word_counts = {'happy': 5, 'because': 3, 'i': 2, 'am': 2, 'learning':3, '.': 1}
for word, freq in word_counts.items():
if freq == f:
print(word)
###Output
_____no_output_____
###Markdown
As mentioned in the videos, if there are many $$ replacements in your train and test set, you may get a very low perplexity even though the model itself wouldn't be very helpful. Here is a sample code showing this unwanted effect.
###Code
# many <unk> low perplexity
training_set = ['i', 'am', 'happy', 'because','i', 'am', 'learning', '.']
training_set_unk = ['i', 'am', '<UNK>', '<UNK>','i', 'am', '<UNK>', '<UNK>']
test_set = ['i', 'am', 'learning']
test_set_unk = ['i', 'am', '<UNK>']
M = len(test_set)
probability = 1
probability_unk = 1
# pre-calculated probabilities
bigram_probabilities = {('i', 'am'): 1.0, ('am', 'happy'): 0.5, ('happy', 'because'): 1.0, ('because', 'i'): 1.0, ('am', 'learning'): 0.5, ('learning', '.'): 1.0}
bigram_probabilities_unk = {('i', 'am'): 1.0, ('am', '<UNK>'): 1.0, ('<UNK>', '<UNK>'): 0.5, ('<UNK>', 'i'): 0.25}
# got through the test set and calculate its bigram probability
for i in range(len(test_set) - 2 + 1):
bigram = tuple(test_set[i: i + 2])
probability = probability * bigram_probabilities[bigram]
bigram_unk = tuple(test_set_unk[i: i + 2])
probability_unk = probability_unk * bigram_probabilities_unk[bigram_unk]
# calculate perplexity for both original test set and test set with <UNK>
perplexity = probability ** (-1 / M)
perplexity_unk = probability_unk ** (-1 / M)
print(f"perplexity for the training set: {perplexity}")
print(f"perplexity for the training set with <UNK>: {perplexity_unk}")
###Output
_____no_output_____
###Markdown
Smoothing Add-k smoothing was described as a method for smoothing of the probabilities for previously unseen n-grams. Here is an example code that shows how to implement add-k smoothing but also highlights a disadvantage of this method. The downside is that n-grams not previously seen in the training dataset get too high probability. In the code output bellow you'll see that a phrase that is in the training set gets the same probability as an unknown phrase.
###Code
def add_k_smooting_probability(k, vocabulary_size, n_gram_count, n_gram_prefix_count):
numerator = n_gram_count + k
denominator = n_gram_prefix_count + k * vocabulary_size
return numerator / denominator
trigram_probabilities = {('i', 'am', 'happy') : 2}
bigram_probabilities = {( 'am', 'happy') : 10}
vocabulary_size = 5
k = 1
probability_known_trigram = add_k_smooting_probability(k, vocabulary_size, trigram_probabilities[('i', 'am', 'happy')],
bigram_probabilities[( 'am', 'happy')])
probability_unknown_trigram = add_k_smooting_probability(k, vocabulary_size, 0, 0)
print(f"probability_known_trigram: {probability_known_trigram}")
print(f"probability_unknown_trigram: {probability_unknown_trigram}")
###Output
_____no_output_____
###Markdown
Back-offBack-off is a model generalization method that leverages information from lower order n-grams in case information about the high order n-grams is missing. For example, if the probability of an trigram is missing, use bigram information and so on.Here you can see an example of a simple back-off technique.
###Code
# pre-calculated probabilities of all types of n-grams
trigram_probabilities = {('i', 'am', 'happy'): 0}
bigram_probabilities = {( 'am', 'happy'): 0.3}
unigram_probabilities = {'happy': 0.4}
# this is the input trigram we need to estimate
trigram = ('are', 'you', 'happy')
# find the last bigram and unigram of the input
bigram = trigram[1: 3]
unigram = trigram[2]
print(f"besides the trigram {trigram} we also use bigram {bigram} and unigram ({unigram})\n")
# 0.4 is used as an example, experimentally found for web-scale corpuses when using the "stupid" back-off
lambda_factor = 0.4
probability_hat_trigram = 0
# search for first non-zero probability starting with trigram
# to generalize this for any order of n-gram hierarchy,
# you could loop through the probability dictionaries instead of if/else cascade
if trigram not in trigram_probabilities or trigram_probabilities[trigram] == 0:
print(f"probability for trigram {trigram} not found")
if bigram not in bigram_probabilities or bigram_probabilities[bigram] == 0:
print(f"probability for bigram {bigram} not found")
if unigram in unigram_probabilities:
print(f"probability for unigram {unigram} found\n")
probability_hat_trigram = lambda_factor * lambda_factor * unigram_probabilities[unigram]
else:
probability_hat_trigram = 0
else:
probability_hat_trigram = lambda_factor * bigram_probabilities[bigram]
else:
probability_hat_trigram = trigram_probabilities[trigram]
print(f"probability for trigram {trigram} estimated as {probability_hat_trigram}")
###Output
_____no_output_____
###Markdown
InterpolationThe other method for using probabilities of lower order n-grams is the interpolation. In this case, you use weighted probabilities of n-grams of all orders every time, not just when high order information is missing. For example, you always combine trigram, bigram and unigram probability. You can see how this in the following code snippet.
###Code
# pre-calculated probabilities of all types of n-grams
trigram_probabilities = {('i', 'am', 'happy'): 0.15}
bigram_probabilities = {( 'am', 'happy'): 0.3}
unigram_probabilities = {'happy': 0.4}
# the weights come from optimization on a validation set
lambda_1 = 0.8
lambda_2 = 0.15
lambda_3 = 0.05
# this is the input trigram we need to estimate
trigram = ('i', 'am', 'happy')
# find the last bigram and unigram of the input
bigram = trigram[1: 3]
unigram = trigram[2]
print(f"besides the trigram {trigram} we also use bigram {bigram} and unigram ({unigram})\n")
# in the production code, you would need to check if the probability n-gram dictionary contains the n-gram
probability_hat_trigram = lambda_1 * trigram_probabilities[trigram]
+ lambda_2 * bigram_probabilities[bigram]
+ lambda_3 * unigram_probabilities[unigram]
print(f"estimated probability of the input trigram {trigram} is {probability_hat_trigram}")
###Output
_____no_output_____ |
[DIR] corporation_list_by_state_2016/corporation_list_by_state_2016.ipynb | ###Markdown
Corporation List By State**Authors:** Patrick Guo Documenting file sizes of Corporations Lists by State in 2016
###Code
import boto3
import numpy as np
import pandas as pd
pd.plotting.register_matplotlib_converters()
import matplotlib.pyplot as plt
%matplotlib inline
from collections import Counter
import statistics
client = boto3.client('s3')
resource = boto3.resource('s3')
my_bucket = resource.Bucket('daanmatchdatafiles')
###Output
_____no_output_____
###Markdown
Files
###Code
companies = 0
###Output
_____no_output_____
###Markdown
Andaman_Nicobar_Islands_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Andaman_Nicobar_Islands_2016.xlsx"
Andaman_Nicobar_Islands_2016 = pd.ExcelFile(path)
print(Andaman_Nicobar_Islands_2016.sheet_names)
# Combine both sheets
Andaman_Nicobar_Islands_2016_1 = Andaman_Nicobar_Islands_2016.parse('Sheet1')
Andaman_Nicobar_Islands_2016_2 = Andaman_Nicobar_Islands_2016.parse('Sheet2')
Andaman_Nicobar_Islands_2016_merged = Andaman_Nicobar_Islands_2016_1.append(Andaman_Nicobar_Islands_2016_2)
# Reset index
Andaman_Nicobar_Islands_2016_merged = Andaman_Nicobar_Islands_2016_merged.reset_index(drop=True)
Andaman_Nicobar_Islands_2016_merged.head()
shape = Andaman_Nicobar_Islands_2016_merged.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (24979, 15)
###Markdown
Andhra_Pradesh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Andhra_Pradesh_2016.xlsx"
Andhra_Pradesh_2016 = pd.ExcelFile(path)
print(Andhra_Pradesh_2016.sheet_names)
Andhra_Pradesh_2016 = Andhra_Pradesh_2016.parse("Sheet3")
Andhra_Pradesh_2016.head()
shape = Andhra_Pradesh_2016.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (24661, 15)
###Markdown
Arunachal_Pradesh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Arunachal_Pradesh_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
Arunachal_Pradesh_2016 = xl.parse("Sheet3")
Arunachal_Pradesh_2016.head()
shape = Arunachal_Pradesh_2016.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (507, 15)
###Markdown
Bihar_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Bihar_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (23550, 15)
###Markdown
Chandigarh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Chandigarh_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (13418, 15)
###Markdown
Chattisgarh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Chattisgarh_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (8413, 15)
###Markdown
Dadar_Nagar_Haveli_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Dadar_Nagar_Haveli_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (472, 15)
###Markdown
Daman_and_Diu_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Daman_and_Diu_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (327, 15)
###Markdown
Goa_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Goa_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (7756, 15)
###Markdown
Gujarat_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Gujarat_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (87613, 15)
###Markdown
Haryana_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Haryana_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Haryana")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (32014, 15)
###Markdown
Himachal_Pradesh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Himachal_Pradesh_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (5200, 15)
###Markdown
Jammu_and_Kashmir_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Jammu_and_Kashmir_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (4426, 15)
###Markdown
Jharkhand_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Jharkhand_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
# Sheet 1 does not provide helpful information
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (10409, 15)
###Markdown
Karnataka_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Karnataka_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (91425, 15)
###Markdown
Kerala_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Kerala_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (41851, 15)
###Markdown
Lakshadweep_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Lakshadweep_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (13, 15)
###Markdown
Madhya Pradesh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Madhya Pradesh_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (32463, 15)
###Markdown
Maharastra_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Maharastra_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (320957, 15)
###Markdown
Manipur_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Manipur_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (477, 15)
###Markdown
Meghalaya_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Meghalaya_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (966, 15)
###Markdown
Mizoram_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Mizoram_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (113, 15)
###Markdown
Nagaland_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Nagaland_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (512, 15)
###Markdown
Odisha_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Odisha_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (20543, 15)
###Markdown
Puducherry_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Puducherry_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (2894, 15)
###Markdown
Punjab_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Punjab_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (27607, 15)
###Markdown
Rajasthan_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Rajasthan_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (49808, 15)
###Markdown
Tamil_Nadu_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Tamil_Nadu_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (124522, 15)
###Markdown
Telangana_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Telangana_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (86553, 15)
###Markdown
Tripura_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Tripura_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (387, 15)
###Markdown
Uttar_Pradesh_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Uttar_Pradesh_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (77818, 15)
###Markdown
Uttarakhand_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/Uttarakhand_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (5598, 15)
###Markdown
West_Bengal_2016
###Code
path = "s3://daanmatchdatafiles/corporation_list_by_state_2016/West_Bengal_2016.xlsx"
xl = pd.ExcelFile(path)
print(xl.sheet_names)
df = xl.parse("Sheet3")
df.head()
shape = df.shape
print("Shape:", shape)
companies += shape[0]
###Output
Shape: (191029, 15)
###Markdown
Total
###Code
print("Number of comapnies:", companies)
###Output
_____no_output_____ |
sandbox/generate_regression_sp-Copy1.ipynb | ###Markdown
Generating Simpson's ParadoxWe have been maually setting, but now we should also be able to generate it more programatically. his notebook will describe how we develop some functions that will be included in the `sp_data_util` package.
###Code
# %load code/env
# standard imports we use throughout the project
import numpy as np
import pandas as pd
import seaborn as sns
import scipy.stats as stats
import matplotlib.pyplot as plt
import mlsim
from mlsim import sp_plot
###Output
_____no_output_____
###Markdown
We have been thinking of SP hrough gaussian mixture data, so we'll first work wih that. To cause SP we need he clusters to have an opposite trend of the per cluster covariance.
###Code
# setup
r_clusters = -.6 # correlation coefficient of clusters
cluster_spread = .8 # pearson correlation of means
p_sp_clusters = .5 # portion of clusters with SP
k = 5 # number of clusters
cluster_size = [2,3]
domain_range = [0, 20, 0, 20]
N = 200 # number of points
p_clusters = [1.0/k]*k
# keep all means in the middle 80%
mu_trim = .2
# sample means
center = [np.mean(domain_range[:2]),np.mean(domain_range[2:])]
mu_transform = np.repeat(np.diff(domain_range)[[0,2]]*(mu_trim),2)
mu_transform[[1,3]] = mu_transform[[1,3]]*-1 # sign flip every other
mu_domain = [d + m_t for d, m_t in zip(domain_range,mu_transform)]
corr = [[1, cluster_spread],[cluster_spread,1]]
d = np.sqrt(np.diag(np.diff(mu_domain)[[0,2]]))
cov = np.dot(d,corr).dot(d)
# sample a lot of means, just for vizualization
# mu = np.asarray([np.random.uniform(*mu_domain[:2],size=k*5), # uniform in x
# np.random.uniform(*mu_domain[2:],size=k*5)]).T # uniform in y
mu = np.random.multivariate_normal(center, cov,k*50)
sns.regplot(mu[:,0], mu[:,1])
plt.axis(domain_range);
# mu
###Output
/home/smb/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:1706: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval
###Markdown
However independent sampling isn't really very uniform and we'd like to ensure the clusters are more spread out, so we can use some post processing to thin out close ones.
###Code
mu_thin = [mu[0]] # keep the first one
p_dist = [1]
# we'll use a gaussian kernel around each to filter and only the closest point matters
dist = lambda mu_c,x: stats.norm.pdf(min(np.sum(np.square(mu_c -x),axis=1)))
for m in mu:
p_keep = 1- dist(mu_thin,m)
if p_keep > .99:
mu_thin.append(m)
p_dist.append(p_keep)
mu_thin = np.asarray(mu_thin)
sns.regplot(mu_thin[:,0], mu_thin[:,1])
plt.axis(domain_range)
###Output
/home/smb/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:1706: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval
###Markdown
Now, we can sample points on top of that, also we'll only use the first k
###Code
sns.regplot(mu_thin[:k,0], mu_thin[:k,1])
plt.axis(domain_range)
###Output
/home/smb/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:1706: FutureWarning: Using a non-tuple sequence for multidimensional indexing is deprecated; use `arr[tuple(seq)]` instead of `arr[seq]`. In the future this will be interpreted as an array index, `arr[np.array(seq)]`, which will result either in an error or a different result.
return np.add.reduce(sorted[indexer] * weights, axis=axis) / sumval
###Markdown
Keeping only a few, we can end up with ones in the center, but if we sort them by the distance to the ones previously selected, we get them spread out a little more
###Code
# sort by distance
mu_sort, p_sort = zip(*sorted(zip(mu_thin,p_dist),
key = lambda x: x[1], reverse =True))
mu_sort = np.asarray(mu_sort)
sns.regplot(mu_sort[:k,0], mu_sort[:k,1])
plt.axis(domain_range)
# cluster covariance
cluster_corr = np.asarray([[1,r_clusters],[r_clusters,1]])
cluster_std = np.diag(np.sqrt(cluster_size))
cluster_cov = np.dot(cluster_std,cluster_corr).dot(cluster_std)
# sample from a GMM
z = np.random.choice(k,N,p_clusters)
x = np.asarray([np.random.multivariate_normal(mu_sort[z_i],cluster_cov) for z_i in z])
# make a dataframe
latent_df = pd.DataFrame(data=x,
columns = ['x1', 'x2'])
# code cluster as color and add it a column to the dataframe
latent_df['color'] = z
sp_plot(latent_df,'x1','x2','color')
###Output
_____no_output_____
###Markdown
We might not want all of the clusters to have the reveral though, so we can also sample the covariances
###Code
# cluster covariance
p_sp_clusters =.8
cluster_size = [4,4]
cluster_std = np.diag(np.sqrt(cluster_size))
cluster_corr_sp = np.asarray([[1,r_clusters],[r_clusters,1]]) # correlation with sp
cluster_cov_sp = np.dot(cluster_std,cluster_corr_sp).dot(cluster_std) #cov with sp
cluster_corr = np.asarray([[1,-r_clusters],[-r_clusters,1]]) #correlation without sp
cluster_cov = np.dot(cluster_std,cluster_corr).dot(cluster_std) #cov wihtout sp
cluster_covs = [cluster_corr_sp, cluster_corr]
# sample the[0,1] k times
c_sp = np.random.choice(2,k,p=[p_sp_clusters,1-p_sp_clusters])
print(c_sp)
# sample from a GMM
z = np.random.choice(k,N,p_clusters)
print(z)
cov_noise = lambda : np.random.permutation([.5*np.random.random(),np.random.random()])
# cluster_covs_all = [cluster_covs[c_i]*np.random.random()/5*(c_i+1) for c_i in c_sp]
cluster_covs_all = [cluster_covs[c_i]*np.random.random()*2*(i+1) for i,c_i in enumerate(c_sp)]
mu_p = [np.random.multivariate_normal(mu,cov) for mu,cov in zip(mu_sort,cluster_covs_all)]
x = np.asarray([np.random.multivariate_normal(mu_sort[z_i],cluster_covs_all[z_i]) for z_i in z])
x2 = np.asarray([np.random.multivariate_normal(mu_p[z_i],cluster_covs_all[z_i]) for z_i in z])
# x = np.asarray([np.random.multivariate_normal(mu_sort[z_i],[[1,.5],[.5,.1]]) for z_i in z])
x = np.concatenate((x,x2),axis=0)
# make a dataframe
latent_df = pd.DataFrame(data=x,
columns = ['x1', 'x2'])
# code cluster as color and add it a column to the dataframe
latent_df['color'] = list(z)*2
sp_plot(latent_df,'x1','x2','color')
b.shape
x.shape
np.random.permutation
cluster_covs[0]*.1
[p_sp_clusters,1-p_sp_clusters]
c_sp
###Output
_____no_output_____
###Markdown
We'll call this construction of SP `geometric_2d_gmm_sp` and it's included in the `sp_data_utils` module now, so it can be called as follows. We'll change the portion of clusters with SP to 1, to ensure that all are SP.
###Code
p_sp_clusters = .9
sp_df2 = mlsim.geometric_2d_gmm_sp(r_clusters,cluster_size,cluster_spread,
p_sp_clusters, domain_range,k,N,p_clusters)
sp_plot(sp_df2,'x1','x2','color')
###Output
_____no_output_____
###Markdown
With this, we can start to see how the parameters control a little
###Code
# setup
r_clusters = -.9 # correlation coefficient of clusters
cluster_spread = .8 # pearson correlation of means
p_sp_clusters = 1 # portion of clusters with SP
k = 5 # number of clusters
cluster_size = [1,5]
domain_range = [0, 20, 0, 20]
N = 200 # number of points
p_clusters = [.5, .2, .1, .1, .1]
sp_df3 = mlsim.geometric_2d_gmm_sp(r_clusters,cluster_size,cluster_spread,
p_sp_clusters, domain_range,k,N,p_clusters)
sp_plot(sp_df3,'x1','x2','color')
sp_df3.head()
plot_df = sp_df3.copy()
def adjust(row):
ad = np.random.rand()*4
row['x1'] += ad + np.random.rand()
row['x2'] -= ad + np.random.rand()
return row
plot_df[plot_df['color']==2] = plot_df[plot_df['color']==2].apply(adjust, axis=1)
plot_df.rename(columns={'color':'name','x1':'alcohol intake','x2':'IQ'},inplace=True)
plot_df.replace({0:'Sam',1:'Carl',2:'Mark',3:'Becky',4:'Mary'},inplace=True)
plot_df.head()
plot_min = plot_df.groupby('name').apply(lambda x: x.sample(10))
sns.lmplot('alcohol intake','IQ' , plot_min, hue='name',ci=None)
# adda whole data regression line, but don't cover the scatter data
sns.regplot('alcohol intake','IQ' , plot_min, color='black', scatter=False,ci=None )
sns.set(font_scale=1.5)
sp_df3.head()
###Output
_____no_output_____
###Markdown
We might want to add multiple views, so we added a function that takes the same parameters or lists to allow each view to have different parameters. We'll look first at just two views with the same parameters, both as one another and as above
###Code
many_sp_df = mlsim.geometric_indep_views_gmm_sp(2,r_clusters,cluster_size,cluster_spread,p_sp_clusters,
domain_range,k,N,p_clusters)
sp_plot(many_sp_df,'x1','x2','A')
sp_plot(many_sp_df,'x3','x4','B')
many_sp_df.head()
###Output
200
4
###Markdown
We can also look at the pairs of variables that we did not design SP into and see that they have vey different structure
###Code
# f, ax_grid = plt.subplots(2,2) # , fig_size=(10,10)
sp_plot(many_sp_df,'x1','x4','A')
sp_plot(many_sp_df,'x2','x4','B')
sp_plot(many_sp_df,'x2','x3','B')
sp_plot(many_sp_df,'x1','x3','B')
###Output
_____no_output_____
###Markdown
And we can set up the views to be different from one another by design
###Code
# setup
r_clusters = [.8, -.2] # correlation coefficient of clusters
cluster_spread = [.8, .2] # pearson correlation of means
p_sp_clusters = [.6, 1] # portion of clusters with SP
k = [5,3] # number of clusters
cluster_size = [4,4]
domain_range = [0, 20, 0, 20]
N = 200 # number of points
p_clusters = [[.5, .2, .1, .1, .1],[1.0/3]*3]
many_sp_df_diff = mlsim.geometric_indep_views_gmm_sp(2,r_clusters,cluster_size,cluster_spread,p_sp_clusters,
domain_range,k,N,p_clusters)
sp_plot(many_sp_df_diff,'x1','x2','A')
sp_plot(many_sp_df_diff,'x3','x4','B')
many_sp_df.head()
###Output
200
4
###Markdown
And we can run our detection algorithm on this as well.
###Code
many_sp_df_diff_result = dsp.detect_simpsons_paradox(many_sp_df_diff)
many_sp_df_diff_result
###Output
_____no_output_____
###Markdown
We designed in SP to occur between attributes `x1` and `x2` with respect to `A` and 2 & 3 in grouby by B, for portions fo the subgroups. We detect other occurences. It can be interesting to exmine trends between the deisnged and spontaneous occurences of SP, so
###Code
designed_SP = [('x1','x2','A'),('x3','x4','B')]
des = []
for i,r in enumerate(many_sp_df_diff_result[['attr1','attr2','groupbyAttr']].values):
if tuple(r) in designed_SP:
des.append(i)
many_sp_df_diff_result['designed'] = 'no'
many_sp_df_diff_result.loc[des,'designed'] = 'yes'
many_sp_df_diff_result.head()
r_clusters = -.9 # correlation coefficient of clusters
cluster_spread = .6 # pearson correlation of means
p_sp_clusters = .5 # portion of clusters with SP
k = 5 # number of clusters
cluster_size = [5,5]
domain_range = [0, 20, 0, 20]
N = 200 # number of points
p_clusters = [1.0/k]*k
many_sp_df_diff = mlsim.geometric_indep_views_gmm_sp(3,r_clusters,cluster_size,cluster_spread,p_sp_clusters,
domain_range,k,N,p_clusters)
sp_plot(many_sp_df_diff,'x1','x2','A')
sp_plot(many_sp_df_diff,'x3','x4','B')
sp_plot(many_sp_df_diff,'x3','x4','A')
many_sp_df_diff.head()
###Output
200
6
|
Mission_to_Mars/Mision_to_Mars.ipynb | ###Markdown
NASA Mars News* Scrape the [NASA Mars News Site](https://mars.nasa.gov/news/) and collect the latest News Title and Paragraph Text. Assign the text to variables that you can reference later.
###Code
#website
url= 'https://mars.nasa.gov/news/'
#visit the website
browser.visit(url)
Title=browser.find_by_css('div.content_title a').text
Title
#html object
html = browser.html
#parse with a beautiful soup
soup = BeautifulSoup(html, 'html.parser')
#browser.find_by_css('div.class_=rollover_description_inner').text
Paragraph = soup.find('div',class_='article_teaser_body').text
Paragraph
###Output
_____no_output_____
###Markdown
JPL Mars Space Images - Featured Image
###Code
#set up Splinter
executable_path = {'executable_path': ChromeDriverManager().install()}
browser = Browser ('chrome', **executable_path, headless = False)
#website
url= 'https://data-class-jpl-space.s3.amazonaws.com/JPL_Space/index.html'
#visit the website
browser.visit(url)
#Find and Save the featured image
browser.click_link_by_partial_text('FULL IMAGE')
browser.find_by_css('img.fancybox-image')['src']
#website
url= 'https://data-class-jpl-space.s3.amazonaws.com/JPL_Space/index.html'
#visit the website
browser.visit(url)
html=browser.html
#create beautifulsoup object with parse
soup = BeautifulSoup(html, 'html.parser')
#get the results and return those as a list
#get the featured image
featured_image_url = soup.find_all('img',class_='headerimage fade-in')
featured_image_url
###Output
_____no_output_____
###Markdown
Mars Facts* Visit the Mars Facts webpage [here](https://space-facts.com/mars/) and use Pandas to scrape the table containing facts about the planet including Diameter, Mass, etc.
###Code
pd.read_html('https://space-facts.com/mars/')[0].to_html()
###Output
_____no_output_____
###Markdown
Mars Hemispheres * Visit the USGS Astrogeology site [here](https://astrogeology.usgs.gov/search/results?q=hemisphere+enhanced&k1=target&v1=Mars) to obtain high resolution images for each of Mar's hemispheres.* You will need to click each of the links to the hemispheres in order to find the image url to the full resolution image.* Save both the image url string for the full resolution hemisphere image, and the Hemisphere title containing the hemisphere name. Use a Python dictionary to store the data using the keys `img_url` and `title`.* Append the dictionary with the image url string and the hemisphere title to a list. This list will contain one dictionary for each hemisphere.```python Example:hemisphere_image_urls = [ {"title": "Valles Marineris Hemisphere", "img_url": "..."}, {"title": "Cerberus Hemisphere", "img_url": "..."}, {"title": "Schiaparelli Hemisphere", "img_url": "..."}, {"title": "Syrtis Major Hemisphere", "img_url": "..."},]```- - -
###Code
url='https://astrogeology.usgs.gov/search/results?q=hemisphere+enhanced&k1=target&v1=Mars'
#visit the website
browser.visit(url)
links = browser.find_by_css('a.itemLink h3')
List=[]
for i in range(len(links)):
hemisphere = {}
hemisphere['title']=browser.find_by_css('a.itemLink h3')[i].text
browser.find_by_css('a.itemLink h3').click()
hemisphere['url']=browser.find_by_text('Sample')['href']
browser.back()
List.append(hemisphere)
browser.quit()
List
###Output
_____no_output_____ |
ml/11-pytorch/tensor-index.ipynb | ###Markdown
tensor的索引操作
###Code
import torch
tensor = torch.arange(2,10)
tensor
print(tensor[3])
print(tensor[2:5])
print(tensor[:4])
print(tensor[-3:])
index = [1,3,5]
print(tensor[index])
###Output
tensor([3, 5, 7])
|
16.NLP/Natural_Language_Processing.ipynb | ###Markdown
Cleaning text
###Code
import re
import nltk
nltk.download('stopwords') # for removing all stopwrods as they does not contain any value
from nltk.corpus import stopwords
from nltk.stem.porter import PorterStemmer # for stemming
corpus = []
for i in range(len(dataset)):
review = re.sub('[^a-zA-Z]', ' ', dataset['Review'][i])
review = review.lower()
review = review.split()
ps = PorterStemmer()
allStopWord = stopwords.words('english')
allStopWord.remove('not')
allStopWord.remove("aren't")
allStopWord.remove("isn't")
review = [ps.stem(word) for word in review if not word in set(allStopWord)]
review = ' '.join(review)
corpus.append(review)
corpus[:4]
###Output
_____no_output_____
###Markdown
Creating a Bag of Words model
###Code
from sklearn.feature_extraction.text import CountVectorizer
cv = CountVectorizer(max_features=1500)
X = cv.fit_transform(corpus).toarray()
y = dataset['Liked']
print(X)
len(X[0])
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=101)
from sklearn.naive_bayes import GaussianNB
classifier = GaussianNB()
classifier.fit(X_train, y_train)
prediction = classifier.predict(X_test)
from sklearn.metrics import confusion_matrix, classification_report
print(confusion_matrix(y_test, prediction))
print(classification_report(y_test, prediction))
from sklearn.ensemble import RandomForestClassifier
classifier = RandomForestClassifier(n_estimators=100)
classifier.fit(X_train, y_train)
prediction = classifier.predict(X_test)
print(confusion_matrix(y_test, prediction))
print(classification_report(y_test, prediction))
###Output
[[87 15]
[18 80]]
precision recall f1-score support
0 0.83 0.85 0.84 102
1 0.84 0.82 0.83 98
accuracy 0.83 200
macro avg 0.84 0.83 0.83 200
weighted avg 0.84 0.83 0.83 200
###Markdown
New Positive review
###Code
def newReview(new_review):
new_review = re.sub('[^a-zA-Z]', ' ', new_review)
new_review = new_review.lower()
new_review = new_review.split()
ps = PorterStemmer()
all_stopwords = stopwords.words('english')
all_stopwords.remove('not')
new_review = [ps.stem(word) for word in new_review if not word in set(all_stopwords)]
new_review = ' '.join(new_review)
new_corpus = [new_review]
new_X_test = cv.transform(new_corpus).toarray()
new_y_pred = classifier.predict(new_X_test)
print(new_y_pred)
newReview('I hate this restaurant so much')
newReview('I love this restaurant so much')
###Output
_____no_output_____ |
notebooks/training/pygeode-virtualenv.ipynb | ###Markdown
Re-runnable virtual environment setupThis notebook is a re-runnable set of instructions to set up a virtual environment and install a package `pygeode` into it.
###Code
# Import the required packages
import os
# Change current working directory to top of the repository: ~/ceda-notebooks/
os.chdir('../..')
from scripts.utils import venv_utils
# Define the name of your venv you wish to create
venv_name = 'venv-notebook'
# List all the packages you want to install into a variable:
packages = [
"pygeode",
]
# Setup the venv to create, activate and install packages
venv_utils.setup_venv(venv_name=venv_name, packages=packages)
import pygeode
?pygeode
###Output
_____no_output_____ |
src/posts/2018-11-26-galton-board/main.ipynb | ###Markdown
title: Simulating a Galton board with Markov chains, eigenvalues and Pythondescription: Use the stationary distribution of an absorbing markov chain to simulate a Galton board
###Code
from IPython.display import set_matplotlib_formats
set_matplotlib_formats('png', 'svg')
###Output
_____no_output_____
###Markdown
The [galton board](https://en.wikipedia.org/wiki/Bean_machine) (also sometimes called a "Bean machine") is a lovely way of visualising the normal distribution.This can be modelled mathematically using [Markov chains](https://en.wikipedia.org/wiki/Markov_chain).We first need to set up the state space we're going to use. Here is a small diagram showing how it will be done:There are a few things there that don't look immediately like the usual galton board:- It is essentially rotated: this is purely to make the mathematical formulation simpler.- There is a probability \\(p\\): this is to play with later, it is taken as the probability of a given bead to fall to the left when hitting an obstacle. In the usual case: \\(p=1/2\\).- We limit our number of rows of obstacles to \\(N\\).Using this, our state space can be written as:\\[ S = \{(i, j) \in \mathbb{R}^2| 0 \leq i \leq N, 0 \leq j \leq i\}\\]A straight forward combinatorial argument gives:\\[ |S| = \binom{N + 1}{2} = \frac{(N + 1)(N + 2)}{2}\\]We are going to use Python throughout this blog post to explore the mathematical objects as we go. First let us create a [python generator](https://medium.freecodecamp.org/how-and-why-you-should-use-python-generators-f6fb56650888) to create our states and also a function to get the size of the state space efficiently:
###Code
import numpy as np
def all_states(number_of_rows):
for i in range(number_of_rows + 1):
for j in range(i + 1):
yield np.array([i, j])
def get_number_of_states(number_of_rows):
return int((number_of_rows + 1) * (number_of_rows + 2) / 2)
list(all_states(number_of_rows=4))
###Output
_____no_output_____
###Markdown
We can test that we have the correct count:
###Code
for number_of_rows in range(10):
assert len(list(all_states(number_of_rows=number_of_rows))) == get_number_of_states(number_of_rows=number_of_rows)
###Output
_____no_output_____
###Markdown
Now that we have our state space, to fully define our Markov chain we need to define our transitions. Given two elements \\(s^{(1)}, s^{(2)}\in S\\) we have:\\[ P_{s^{(1)}, s^{(2)}} = \begin{cases} p & \text{ if }s^{(2)} - s^{(1)} = (1, 0)\text{ and }{s_1^{(2)}} < N \\ 1 - p & \text{ if }s^{(2)} - s^{(1)} = (1, 1)\text{ and }{s_1^{(2)}} < N \\ 1 & \text{ if }s^{(2)} = N\\ 0 & \text{otherwise} \\ \end{cases}\\]The first two values ensure the beads bounce to the left and right accordingly and the third line ensures the final row is absorbing (the beads stay put once they've hit the bottom).Here's some Python code that replicates this:
###Code
def transition_probability(in_state, out_state, number_of_rows, probability_of_falling_left):
if in_state[0] == number_of_rows and np.array_equal(in_state, out_state):
return 1
if np.array_equal(out_state - in_state, np.array([1, 0])):
return probability_of_falling_left
if np.array_equal(out_state - in_state, np.array([1, 1])):
return 1 - probability_of_falling_left
return 0
###Output
_____no_output_____
###Markdown
We can use the above and the code to get all states to create the transition matrix:
###Code
def get_transition_matrix(number_of_rows, probability_of_falling_left):
number_of_states = get_number_of_states(number_of_rows=number_of_rows)
P = np.zeros((number_of_states, number_of_states))
for row, in_state in enumerate(all_states(number_of_rows=number_of_rows)):
for col, out_state in enumerate(all_states(number_of_rows=number_of_rows)):
P[row, col] = transition_probability(in_state, out_state, number_of_rows, probability_of_falling_left)
return P
P = get_transition_matrix(number_of_rows=3, probability_of_falling_left=1/2)
P
###Output
_____no_output_____
###Markdown
Once we have done this we are more of less there. With \\(\pi=(1,...0)\\), the product: \\(\pi P\\) gives the distribution of our beads after they drop past the first row. Thus the following gives us the final distribution over all the states after the beads get to the last row:\\[ \pi P ^ N\\]Here is some Python code that does exactly this:
###Code
def expected_bean_drop(number_of_rows, probability_of_falling_left):
number_of_states = get_number_of_states(number_of_rows)
pi = np.zeros(number_of_states)
pi[0] = 1
P = get_transition_matrix(number_of_rows=number_of_rows, probability_of_falling_left=probability_of_falling_left)
return (pi @ np.linalg.matrix_power(P, number_of_rows))[-(number_of_rows + 1):]
expected_bean_drop(number_of_rows=3, probability_of_falling_left=1/2)
###Output
_____no_output_____
###Markdown
We can then plot this (using `matplotlib`):
###Code
import matplotlib.pyplot as plt
def plot_bean_drop(number_of_rows, probability_of_falling_left, label=None, color=None):
plt.plot(expected_bean_drop(number_of_rows=number_of_rows,
probability_of_falling_left=probability_of_falling_left),
label=label, color=color)
return plt
###Output
_____no_output_____
###Markdown
Let us use this to plot over \\(N=30\\) rows and also overlay with the normal pdf:
###Code
from scipy.stats import norm
plt.figure()
plot_bean_drop(number_of_rows=30, probability_of_falling_left=1 / 2)
plt.ylabel("Probability")
plt.xlabel("Bin position");
###Output
_____no_output_____
###Markdown
We can also plot this for a changing value of \\(p\\):
###Code
from matplotlib import cm
number_of_rows = 30
plt.figure()
for p in np.linspace(0, 1, 10):
plot_bean_drop(
number_of_rows=number_of_rows,
probability_of_falling_left=p,
label=f"p={p:0.02f}",
color=cm.viridis(p)
)
plt.ylabel("Probability")
plt.xlabel("Bin position")
plt.legend();
###Output
_____no_output_____
###Markdown
We see that as the probability of falling left moves from 0 to 1 the distribution slow shifts across the bins.
###Code
import pathlib
p = pathlib.Path("./gif/")
p.mkdir(exist_ok=True)
for p in np.linspace(0, 1, 200):
title = f"{p:0.02f}"
plt.figure()
plot_bean_drop(
number_of_rows=number_of_rows,
probability_of_falling_left=p,
color=cm.viridis(p)
)
plt.title(f"Probability of falling left: $p={title}$")
plt.ylabel("Probability")
plt.xlabel("Bin position")
plt.savefig(f"./gif/{title}.pdf");
###Output
/home/vince/anaconda3/lib/python3.6/site-packages/matplotlib/pyplot.py:537: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`).
max_open_warning, RuntimeWarning)
|
Wi20_content/chan_vese_segmentation.ipynb | ###Markdown
Many segmentation algorithms rely on the existence of clear edges. But what happens when there are no clear edges? Can segmentation still be performed? You perhaps can attenpt a segmentation by performing some smoothing first, but these aren't guaranteed to be successful. Take the following image, for instance:
###Code
import numpy as np
from skimage.segmentation import chan_vese
from skimage.morphology import disk, binary_closing, binary_opening
from skimage.filters import rank
from skimage.data import camera
import matplotlib.pyplot as plt
from skimage.util import img_as_float
from scipy.ndimage.morphology import binary_fill_holes
#im = plt.imread('https://i.stack.imgur.com/xHJHC.png')
#im = camera()
im = plt.imread('https://www.researchgate.net/profile/Peter_Bankhead/publication/260261544/figure/fig30/AS:669432326135812@1536616512124/A-demonstration-that-Poisson-noise-changes-throughout-an-image-a-Part-of-a-spinning.ppm')
im = im[:, :, 0]
plt.imshow(im, cmap='gray')
plt.axis('off')
im.shape
# cv = chan_vese(im, mu=0.25, lambda1=1.0, lambda2=1.5, tol=1e-3, max_iter=250, dt=0.5,
# init_level_set='checkerboard', extended_output=True)
cv = chan_vese(im, mu=0.2, lambda1=1.0, lambda2=1.0, tol=5e-4, max_iter=4000, dt=0.5,
init_level_set='checkerboard', extended_output=True)
fig, axes = plt.subplots(2,2, figsize=(10,10))
axes = axes.flatten()
titles = ['Original Image', 'Chan-Vese Segmentation', 'Final Level Set', 'Evolution of Energy']
C = binary_fill_holes(np.abs(cv[1]) < 0.01)
axes[0].imshow(im, cmap='gray')
axes[1].imshow(cv[0], cmap='gray')
axes[2].imshow(cv[1], cmap='gray')
axes[3].imshow(cv[0] | C, cmap='gray')
#axes[3].plot(cv[2])
#axes[3].imshow(img_as_float(im) - cv[1], cmap='gray')
for i, ax in enumerate(axes):
ax.axis('off')
###Output
_____no_output_____ |
tutorial/09_Functions.ipynb | ###Markdown
Functions 1) IntroductionFunctions are very useful to apply multiple time a block of instructions. Software written with such sets of instructions is also more readable and easy to debug.A function takes argument in input, apply a block of instruction on these, and return a result. 2) DefinitionTo define a new function in Python, one just need tto use the keyword def, for instance:
###Code
def square(x):
return x**2
print(square(2))
###Output
_____no_output_____
###Markdown
As for the loops and the tests, in Python a new function is defined in an indented block of instruction after "def funcname():". Functions can be used to instentiate variables:
###Code
k=square(3)
print(k)
###Output
_____no_output_____
###Markdown
It is also possible to define functions that do not take any argument and do not return anything. For example:
###Code
def hello():
print("hello")
hello()
var=hello()
print(var)
###Output
_____no_output_____
###Markdown
The number of arguments taken by the function is free. It is not necessary to type the argument used by a function. This will work as long as the operations performed on the arguments are accepted by the language. For instance let's define the multiply function:
###Code
def multiply(x,y):
return x*y
print(multiply(2,3))
print(multiply(2.,3.))
print(multiply(2,"three"))
print(multiply("two","three"))
###Output
_____no_output_____
###Markdown
It is possible to return multiple values at the same time. For instance on cae use tuple (we will see this later) or lists:
###Code
def square_cube(x):
return x**2 , x**3
def square_cube2(x):
return [x**2,x**3]
ntuple=square_cube(3)
print(ntuple,type(ntuple))
lists=square_cube2(3)
print(lists,type(lists))
###Output
_____no_output_____
###Markdown
One can use these functions to instentiate multiple values ate the same time, for instance:
###Code
z1,z2=square_cube2(3)
print(z1,z2)
###Output
_____no_output_____
###Markdown
2) ArgumentsWhat happen if we define a function that expects two arguments, and we use this function only with a single number:
###Code
def multiply(x,y):
return x*y
print(multiply(2,3))
print(multiply(2))
###Output
_____no_output_____
###Markdown
Python return an error, because the positional arguments that were used in the function declaration were not provided correctly. Note that it is mandatory to use the arguments defined in the function in the proper order. It is possible to define a default assignement value to the variables used, or keyword argument:
###Code
def ret(x=1):
return x
print(ret())
print(ret(10))
###Output
_____no_output_____
###Markdown
This can be done for multiple arguments:
###Code
def fct (x=0, y=0, z =0):
return x, y, z
print(fct())
print(fct(10))
print(fct(10,8))
print(fct(10,8,3))
###Output
_____no_output_____
###Markdown
It is possible to change the order of the keyword arguments when a function is called if one explicitely give the name of the variable:
###Code
print(fct (z=10 , x=3, y =80))
print(fct (z=10 , y =80))
###Output
_____no_output_____
###Markdown
It is possible to mix definitions with positionnal and keywords arguments, but the positionnal arguments have to be declared first:
###Code
def fct(a, b, x=0, y=0, z =0):
return a, b, x, y, z
print(fct(1,1))
###Output
_____no_output_____
###Markdown
3) local and global variablesA variable is local when it is created inside a function. A variable is global when it is created in the main program. For instance:
###Code
# functions definition
def square (x):
y = x **2
return y
# Main program
z = 5
result = square (z)
print ( result )
print(y)
###Output
_____no_output_____
###Markdown
In this code, x and y are local variables because they were used to defined the square function, while z and results are global variables. If one try to access y in the main program, Python return a crash. It is however possible to use global variable in functions:
###Code
# functions definition
def func():
print(x)
# Main program
x = 3
func()
###Output
_____no_output_____
###Markdown
It is not possible to modify a global variable in a function:
###Code
# functions definition
def func2():
x2+=1
# Main program
x2 = 3
func2()
x2
###Output
_____no_output_____
###Markdown
Except if one uses the global word:
###Code
# functions definition
def func3():
global x3
x3+=1
# Main program
x3 = 3
func3()
x3
###Output
_____no_output_____
###Markdown
In general, it is not very wise to use global functions across different functions, since the code will be harder to read and to debug. 4) Calling a function from another functionIt is possible to call a function from another function, if the first function has been loaded by Python. For instance:
###Code
# functions definition
def polynom (x):
return (x **2 - 2*x + 1)
def calc_vals (beg , end ):
list_vals = []
for x in range (beg , end + 1):
list_vals . append ( polynom (x))
return list_vals
# Main program
print ( calc_vals (-5, 5))
###Output
_____no_output_____
###Markdown
A function can even call itself, such a function will be called a recusive function. For instance if one need to define the mathematical operator Factorial: $n!=n \times n-1 \times ... \times 2 \times 1$, this can be obtained using the following lines:
###Code
# functions definition
def factorial (n):
if n == 1:
return 1
else :
return n * factorial(n - 1)
# Main program
print ( factorial(4))
###Output
_____no_output_____
###Markdown
In summary the function call itself until the moment where n is 1, and decrease n by one at each iteration. 5) Functions and modifiable types:When using modifiable types such as lists in functions one need to pay attention. For instance:
###Code
def func():
list[1] = -127
print(id(list))
# Main program
list = [1 ,2 ,3]
print(id(list))
func()
list
###Output
4591805192
4591805192
###Markdown
Modify the list. One can observe that the addresses of the two variables displayed using the ``id()`` function return the same number. By default if a list is passed as argument to a function, it is its reference that is passed to the function, and therefore the list can be modified:
###Code
def func(x):
x[1] = -15
# Main program
list = [1 ,2 ,3]
func(list)
list
###Output
_____no_output_____
###Markdown
If one want to avoid this behavior, one need to pass the list explicitely:
###Code
def func(x):
x[1] = -15
# Main program
list = [1 ,2 ,3]
func(list[:])
list
###Output
_____no_output_____ |
ex10_Utilisation_des_vues_pour_simplifier_les_requêtes.ipynb | ###Markdown
ex10 - Utilisation des vues pour simplifier les requêtesL'un des beaux aspects du modèle de données relationnel et SQL est que la sortie d'une requête est également une table, une relation pour être précis. Il peut s'agir d'une seule colonne ou d'une seule ligne, mais il s'agit néanmoins d'un tableau. Une vue est une requête qui peut être utilisée comme une table. Une vue peut être considérée comme une table virtuelle qui ne contient pas de données. Elle correspond juste à une requête. Chaque fois qu'une vue est accédée, la requête sous-jacente est exécutée et les résultats renvoyés peuvent être utilisés comme s'ils constituaient une table réelle.Il y a plusieurs raisons (http://www.sqlitetutorial.net/sqlite-create-view/) pour utiliser les vues. Gardez à l’esprit le principe de programmation DRY: ne vous répétez pas. Éviter les répétitions permet de gagner du temps et d'éviter les erreurs inutiles. C'est l'une des bonnes raisons pour lesquelles nous enregistrons les requêtes sous forme de vues de base de données réutilisables.Les vues SQLite sont créées à l'aide de l'instruction CREATE VIEW. Les vues SQLite peuvent être créées à partir d'une seule table, de plusieurs tables ou d'une autre vue. Voici la syntaxe de la commande ***CREATE VIEW*** de base (http://www.sqlitetutorial.net/sqlite-create-view/):>CREATE [TEMP | TEMPORARY] VIEW view_name AS>SELECT column1, column2.....>FROM table_name>WHERE [condition];La vue dans SQLite est en lecture seule. Cela signifie que vous ne pouvez pas utiliser les instructions INSERT, DELETE et UPDATE pour mettre à jour les données dans les tables de base via la vue.
###Code
%load_ext sql
from google.colab import drive
# drive.mount('/content/gdrive')
drive.mount("/content/gdrive", force_remount=True)
###Output
Mounted at /content/gdrive
###Markdown
1. Connection à la database demo.db3
###Code
%sql sqlite:////content/gdrive/MyDrive/Partage/Notebooks_Serie_1/demo.db3
###Output
_____no_output_____
###Markdown
Si vous ne vous souvenez pas des tables présentes dans la database de démonstration, vous pouvez toujours utiliser la commande suivante pour les retrouver.
###Code
%sql SELECT name FROM sqlite_master WHERE type='table'
###Output
* sqlite:////content/gdrive/MyDrive/Partage/Notebooks_Serie_1/demo.db3
Done.
###Markdown
2. Simplifier les requêtes avec l'utilisation de vuesDans le Notebook précédent, nous avons utilisé CASE et Subquery pour calculer le ruissellement saisonnier à partir du tableau de rch. Ici, nous utiliserons une vue pour simplifier le calcul. 2.1 Rappel sur le calcul du ruissellement saisonnier
###Code
%%sql sqlite://
SELECT RCH, Quarter, AVG(FLOW_OUTcms) as Runoff
FROM(
SELECT RCH, YR,
CASE
WHEN (MO) BETWEEN 3 AND 5 THEN 'MAM'
WHEN (MO) BETWEEN 6 and 8 THEN 'JJA'
WHEN (MO) BETWEEN 9 and 11 THEN 'SON'
ELSE 'DJF'
END Quarter,
FLOW_OUTcms
from rch)
GROUP BY RCH, Quarter
LIMIT 5
###Output
Done.
###Markdown
2.2 Création d'une view
###Code
%%sql sqlite://
CREATE VIEW RCH_VW AS SELECT RCH, YR,
CASE
WHEN (MO) BETWEEN 3 AND 5 THEN 'MAM'
WHEN (MO) BETWEEN 6 and 8 THEN 'JJA'
WHEN (MO) BETWEEN 9 and 11 THEN 'SON'
ELSE 'DJF'
END Quarter,
FLOW_OUTcms
from rch
###Output
Done.
###Markdown
Requêtons maintenant la vue SSN_RCH
###Code
%%sql sqlite://
SELECT *
FROM RCH_VW
LIMIT 5
###Output
Done.
###Markdown
2.3 Recalculons le ruissellement avec la vue :Le code est réellement simplifié et plus court
###Code
%%sql sqlite://
SELECT RCH, Quarter, AVG(FLOW_OUTcms) as Runoff
FROM RCH_VW
GROUP BY RCH, Quarter
LIMIT 5
###Output
Done.
###Markdown
2.4 Suppression de vuesIl est de plus assez facile de supprimer des vues :
###Code
%sql DROP VIEW RCH_VW
###Output
* sqlite:////content/gdrive/MyDrive/Partage/Notebooks_Serie_1/demo.db3
Done.
|
notebooks/04_exercises_without_sol.ipynb | ###Markdown
Exercises for workshop 4 Even though we do not have a strong focus on mathematics in this workshop series, we encourage you to get familiar with some mathematical topics since machine learning in general is highly depending on them. If you want to you can use the following materials provided by the Standford university to review or refresh your mathematical knowledge:- Linear algebra http://cs229.stanford.edu/section/cs229-linalg.pdf- Probability theory http://cs229.stanford.edu/summer2020/cs229-prob.pdf
###Code
import math as m
import numpy as np
###Output
_____no_output_____
###Markdown
We have 10 samples from the cities Berlin, Munich, Stuttgart, Nuremberg, Hamburg, Hannover, Augsburg, Halle, Fürth and Ingolstadt (for an actual good model that might not be enough, but it suffices for showing the concepts).To make things easier we include the bias term in our feature array by adding a 1 to every sample of longitude and latitude data. We get the following feature, label and weight arrays:
###Code
features = np.array([[1, 52.5167, 13.3833], [1, 48.1372, 11.5755], [1, 48.7761, 9.1775], [1, 49.4539, 11.0775],
[1, 53.55, 10], [1, 52.3744, 9.7386], [1, 48.3717, 10.8983], [1, 51.4828, 11.9697],
[1, 49.4783, 10.9903], [1, 48.7636, 11.4261]]) # [biasTerm, lat, lng]
labels = np.array([0, 1, 0, 1, 0, 0, 1, 0, 1, 1]) # is in Bavaria
weights = np.array([1,.5,.5])
def sigmoid(z):
return 1.0 /(1 + np.exp(-z))
def predict(features, weights):
return sigmoid(np.dot(features, weights))
predict(features, weights)
###Output
_____no_output_____
###Markdown
Task 1:You are given the mentioned parametric function, the sigmoid function, which outputs values between 0 and 1. It is thus suitable for a simple classifier by providing the probability of belonging to a certain class. You can check out how it looks on wikipedia.$$ \sigma(z_i)=\frac{1}{1+e^{-z_i}} $$with $ z_i = w_0 + w_1\cdot x_{i,lat} + w_2\cdot x_{i,lng}$Compute the binary cross entropy loss for every sample. It is defined as following:$$ l(\hat{y}_i, y_i) = -(y_i \cdot log(\hat{y}_i) + (1-y_i)\cdot log(1-\hat{y}_i)) $$
###Code
def BCE(features, weights, labels):
# TODO
return pass
BCE(features, weights, labels)
def EmpiricalLoss(features, weights, labels):
N = len(features)
BiCrEn = BCE(features, weights, labels)
loss = BiCrEn.sum() / N
return loss
EmpiricalLoss(features, weights, labels)
###Output
_____no_output_____
###Markdown
Task 2:Compute the gradient of the empirical loss (also called cost function) with the binarycross entropy loss function with regards to the weight vectorw.The gradient for a function $f: \mathbb{R}^n \rightarrow \mathbb{R}$, is defined as follows:$$\nabla f = (\frac{\delta f}{\delta x_1}, ..., \frac{\delta f}{\delta x_n}) $$
###Code
def Gradient(features,weights,labels):
# TODO
return pass
Gradient(features, weights, labels)
###Output
_____no_output_____
###Markdown
Task 3:Gradient descent update steps:$$g^{(t)} = \nabla L(X,Y;w^{(t-1)})$$ $$w^{(t)} = w^{(t-1)} - \eta \cdot g^{(t)}$$Implement the gradient descent method and execute it with a learning rate of $\eta = 0.001$ and about 3000 iterations
###Code
def GradientDescent(eta, features, labels, weights, iterations):
# TODO
return pass
updated_weights = GradientDescent(0.001, features, labels, weights, 3001)
updated_weights
###Output
iteration: 0 loss: 15.449124167438217
iteration: 1000 loss: 0.6036136116977191
iteration: 2000 loss: 0.5981876720530922
iteration: 3000 loss: 0.5950049075701982
###Markdown
Task 4:In the following we want to test how good our small modell is performing. For that we need a classifier which determines acording to our probalities given by our modell if a city is located in bavaria. After implementing the classifier compute the accuracy of your training data and the given test data. You can also take a look at the actual probabillities. Did the empirical loss improve?
###Code
# test data: Würzburg, Rostock, Trier, Rosenheim, Regensburg
test_features = np.array([[1, 49.7944, 9.9294], [1, 54.0833, 12.1333], [1, 49.7567, 6.6414], [1, 47.8561, 12.1289],
[1, 49.0167, 12.0833]]) # [biasTerm, lat, lng]
test_labels = np.array([1, 0, 0, 1, 1]) # is in Bavaria
def classify(features, weights):
# TODO
return pass
# Accuracy and probabilities of training data
# TODO
# Empirical loss change
# TODO
# Accuracy and probabilities of test data
# TODO
###Output
_____no_output_____ |
demos/presidential-elections/Elections.ipynb | ###Markdown
Presidential Election NotebookThis notebook takes the raw Presidential Elections spreadsheet in elections.csv and the electoral college sheet in electoral_college.csv and converts them into a number of tables suitable for the presidential elections dashboard. This involves:1. Splitting the combined field \ - \ into two fields, candidate and party2. Converting years to integers and putting in the missing years (converting '2016', '', '' to 2016, 2016, 20163. Collecting the cells of a particular state and year into a structure, with the individual candidates as a list4. Converting the votes into integers, and then, for each result, adding a percentage float5. generating the individual records (state, year, candidate, party, votes, percentage) as a list6. Creating the data table as a Galyleo Table and sending it to the dashboard. This will form the basis of the Candidate Votes by State and Year and Party Percent by State and Year Charts7. Selecting the rows of this table for Nationwide results. This will be the basis of the pie chart for national share of the vote. Send this to the dashboard.8. Forming a pivot table of percentage of the vote for each party, by state and year. This will form the basis of the Vote history line chart. Send this to the dashboard9. Converting the pivot table to a margin table, which will form the basis of the colored map. Send this to the dashboard.10. Finally, reading in the electoral college results from the CSV file and turning this into a set of records (Year, EV-Democrat, EV-Republican, EV-Other). Send this to the dashboard.All in, we compute five tables to send to the dashboard. These will be filtered using widgets on the dashboard to form six graphs, which respond to the filters to show results for a specific state and year. Step 0: read in the table from the CSV file. After this, the data will be in the variable rows.
###Code
import csv
f = open('elections.csv', 'r')
election_reader = csv.reader(f)
rows = [row for row in election_reader]
f.close()
###Output
_____no_output_____
###Markdown
In the raw data, the candidate field is - . Parse into pairs a dictionary {"name": , "total": }. If the string is "Total", then the party and candidate are both "Total". If there is no dash, there is no name and party is "Other".
###Code
def clean_candidate(raw):
if (raw == 'Total'): return {"Name": "Total", "Party": "Total"}
else:
parsed = raw.split(' - ')
return {"Name": parsed[0], "Party": parsed[1]} if len(parsed) == 2 else {"Name": "", "Party": "Other"}
candidates = [clean_candidate(entry) for entry in rows[1][1:]]
candidates
###Output
_____no_output_____
###Markdown
Parties have gone by various aliases throughout the years; moreover, our dataset goes back to 1828, but the Republican party wasn't formed until 1854 from the Whig Party, which was itself a descendant of the Federalists. As a result, we consolidate parties using this function, and then make sure that every candidate's party is canonized.
###Code
def canonical_name(party):
party_aliases = {'National Republican': "Republican", 'National Union (Republican)': "Republican", 'Whig': "Republican",
'Liberal Republican/Democratic': "Democratic", '(Northern) Democratic': "Democratic",
'Progressive "Bull Moose"': 'Progressive'}
return party_aliases[party] if party in party_aliases else party
for candidate in candidates: candidate["Party"] = canonical_name(candidate["Party"])
###Output
_____no_output_____
###Markdown
The years are blank except for the first column in every group, leading to the following messy bit of code to assign a year to every record
###Code
class YearCanonizer:
def __init__(self, years):
self.years = [self.canonize_year(year)for year in years]
def canonize_year(self, year):
if (year != ""): self.prev_year = int(year)
return self.prev_year
canonizer = YearCanonizer(rows[0][1:])
for i in range(len(candidates)): candidates[i]["Year"] = canonizer.years[i]
###Output
_____no_output_____
###Markdown
Code which, from the row for a state, and the records {"Name", "Party", "Year"} computes {{"Name", "Party", "Year", "State", "Votes"}, using the fact that the votes are in the same order as the candidates
###Code
# First, convert a string which may be blank or contain commas to a number
def compute_int_from_delimited_string(string):
string = string.strip()
string = string.replace(',', '')
return int(string) if len(string) > 0 else 0
def compute_state_record(state_row):
state_name = state_row[0]
votes = state_row[1:]
result = [candidate.copy() for candidate in candidates]
for candidate in result: candidate["State"] = state_name
for i in range(len(result)): result[i]["Votes"] = compute_int_from_delimited_string(votes[i])
return result
state_records = [compute_state_record(row) for row in rows[2:]]
state_list = []
for record in state_records: state_list.extend(record)
###Output
_____no_output_____
###Markdown
Trim the records with 0 votes and then get the totals for each state and year
###Code
state_list = [record for record in state_list if record["Votes"] > 0]
total_records = [record for record in state_list if record["Name"] == "Total"]
party_records = [record for record in state_list if record["Name"] != "Total"]
totals = {}
for record in total_records: totals[(record["State"], record["Year"])] = record["Votes"]
###Output
_____no_output_____
###Markdown
Compute the percentages
###Code
for record in party_records: record["Percentage"] = 100* record["Votes"]/totals[(record["State"], record["Year"])]
###Output
_____no_output_____
###Markdown
Create first table
###Code
from galyleo.galyleo_table import GalyleoTable
from galyleo.galyleo_constants import GALYLEO_STRING, GALYLEO_NUMBER
table = GalyleoTable("presidential_vote")
schema = [("Year", GALYLEO_NUMBER), ("State", GALYLEO_STRING), ("Name", GALYLEO_STRING), ("Party", GALYLEO_STRING), ("Votes", GALYLEO_NUMBER), ("Percentage", GALYLEO_NUMBER)]
data = [[record["Year"], record["State"], record["Name"], record["Party"], record["Votes"], record["Percentage"]] for record in party_records]
table.load_from_schema_and_data(schema, data)
###Output
_____no_output_____
###Markdown
Send the first table to the dashboard
###Code
from galyleo.galyleo_jupyterlab_client import GalyleoClient
client = GalyleoClient()
client.send_data_to_dashboard(table)
###Output
_____no_output_____
###Markdown
A filtered table, nationwide vote only -- this will drive a pie chart with the national percentage of the vote.
###Code
print('Hello, world')
nationwide_records = [[record[0], record[3], record[5]] for record in data if record[1] == "Nationwide"]
nationwide_schema = [("Year", GALYLEO_NUMBER), ("Party", GALYLEO_STRING), ("Percentage", GALYLEO_NUMBER)]
table = GalyleoTable("nationwide_vote")
table.load_from_schema_and_data(nationwide_schema, nationwide_records)
client.send_data_to_dashboard(table)
###Output
_____no_output_____
###Markdown
Time to form the pivot and margin tables, which we will use for the map and the history graph. One note is that there have been a _lot_ of parties in American history; the Cook database shows 26, and even after we have removed 7 as aliases, above, this leaves 19. This makes for a busy history chart. So what we will do here is choose a party list, and everything else becomes "Other". The party list is a matter of taste; it will of course include Republican and Democrat, but the remainder are personal preference. I'm using "Progressive", "Socialist", and "Reform", since they showed well in 2 or more elections and/or captured 20% of the national vote in one
###Code
parties = ['Democratic', 'Republican', 'Progressive', 'Socialist', 'Reform']
###Output
_____no_output_____
###Markdown
Pivot on percentage to break out by party. The idea here is to create records of the form {State, Year, P1,...,Pn} where each Pi is the name of a party and the value is the percentage of the vote
###Code
# A function which creates an empty pivot record for state and year
def pivot_record(state, year):
result = {"State": state, "Year": year}
for party in parties: result[party] = 0
result["Other"] = 0
return result
# Compute the pivot table. For each record in stripped_list, add the vote to the entry for state and year. If
# none exists, create the record first
pivot_table = {}
party_set = set(parties)
for record in party_records:
if not ((record["State"], record["Year"]) in pivot_table):
pivot_table[(record["State"], record["Year"])] = pivot_record(record["State"], record["Year"])
if record["Party"] in party_set:
pivot_table[(record["State"], record["Year"])][record["Party"]] = record["Percentage"]
else:
pivot_table[(record["State"], record["Year"])]["Other"] = max(record["Percentage"], pivot_table[(record["State"], record["Year"])]["Other"])
###Output
_____no_output_____
###Markdown
The pivot table is now complete. Just prepare the table and send it to the dashboard. Add "Other" to the list of parties, then form the Schema ("State" is a string, everything else is a number), create the table, load it with the data, and send to a dashboard.
###Code
parties.append("Other")
pivot_schema = [("State", GALYLEO_STRING), ("Year", GALYLEO_NUMBER)] + [(party, GALYLEO_NUMBER) for party in parties]
pivot_data = [[record["State"], record["Year"]] + [record[party] for party in parties] for record in pivot_table.values()]
galyleo_pivot_table = GalyleoTable("presidential_vote_history")
galyleo_pivot_table.load_from_schema_and_data(pivot_schema, pivot_data)
client.send_data_to_dashboard(galyleo_pivot_table)
###Output
_____no_output_____
###Markdown
Prepare the margin table. This is going to drive a red/blue/green map, where a Democratic victory is going to be on the scale 5-10, Republican on the scale -5 to -10, and "Other" will be 0. We only need three parties for this one, Democratic, Republican, and Other, so we consolidate the margin table down to a list of length 3
###Code
from functools import reduce
def consolidate(pivot_record):
other = pivot_record[parties[2]]
for party in parties[3:]: other = max(other, pivot_record[party])
result = {"Other": other}
for field in ["State", "Year", "Democratic", "Republican"]: result[field] = pivot_record[field]
return result
margin_party_records = [consolidate(pivot_record) for pivot_record in pivot_table.values()]
###Output
_____no_output_____
###Markdown
Now we have the margin records, and we want to convert them into 5-10 (state lightly to heavily Democratic) (-10 - -5) (state heavily to lightly Republican), and 0 (other). We'll just use a linear scale, capped at 10, so 0-2% is light, 2-4% is next, and 10% or above is heavy. This is adjustable.
###Code
def compute_margin(margin_party_record):
if (margin_party_record["Other"] > max(margin_party_record["Republican"], margin_party_record["Democratic"])): return 0
multiplier = 1 if margin_party_record["Democratic"] > margin_party_record["Republican"] else -1
raw_margin = min(round(abs(margin_party_record["Democratic"] - margin_party_record["Republican"])/2), 5)
return multiplier * (5 + raw_margin)
margins = [[record["State"], record["Year"], compute_margin(record)] for record in margin_party_records]
margin_table = GalyleoTable('presidential_margins')
margin_table.load_from_schema_and_data([('State', GALYLEO_STRING), ('Year', GALYLEO_NUMBER), ('Margin', GALYLEO_NUMBER)], margins)
client.send_data_to_dashboard(margin_table)
###Output
_____no_output_____
###Markdown
Finally, get the electoral college. This is very simple. Just create one record per year, with three fields: Republican, Democratic, Other
###Code
ec_aliases = {'Republican': {'National Republican', 'Whig', 'Republican'}, 'Democratic': {'Democratic/Liberal Republican', 'Independent-Democratic', 'Democratic', 'Democrat'}}
f = open('electoral_college.csv', 'r')
ec_reader = csv.reader(f)
rows = [row for row in ec_reader]
f.close()
class EC_Record:
def __init__(self, year):
self.year = int(year)
self.republican = 0
self.democratic = 0
self.other = 0
def add_record(self, record):
value = int(record[3])
if (record[2] in ec_aliases['Republican']): self.republican = self.republican + value
elif (record[2] in ec_aliases['Democratic']): self.democratic = self.democratic + value
else: self.other = self.other + value
def as_list(self):
return [self.year, self.democratic, self.republican, self.other]
def __repr__(self):
l1 = self.as_list()
return ', '.join([str(elt) for elt in l1])
ec_records = {}
for row in rows[1:]:
year = row[0]
if year not in ec_records: ec_records[year] = EC_Record(year)
ec_records[year].add_record(row)
###Output
_____no_output_____
###Markdown
Load the Electoral College records into a table and send it to the dashboard
###Code
schema = [("Year", GALYLEO_NUMBER), ("Democratic", GALYLEO_NUMBER), ("Republican", GALYLEO_NUMBER), ("Other", GALYLEO_NUMBER)]
ec_table = GalyleoTable("electoral_college")
ec_table.load_from_schema_and_data(schema, [record.as_list() for record in ec_records.values()])
client.send_data_to_dashboard(ec_table)
###Output
_____no_output_____ |
scripts/pub_test/identify_theory_games.ipynb | ###Markdown
Find deviations from Lichess masters DB
###Code
import os
import time
import requests
from dotenv import load_dotenv
load_dotenv()
token = os.environ.get("LICHESS_TOKEN")
import requests
import chess
import chess.pgn
import time
import sys
import berserk
import io
from IPython.display import SVG
from tqdm import tqdm
session = berserk.TokenSession(token)
client = berserk.Client(session)
# Check FEN against Lichess masters DB
def check_fen_against_mastersdb(fen):
FENdict = {}
cachedfen = FENdict.get(fen)
if cachedfen:
r = cachedfen
else:
while True:
payload = {'fen': fen, 'topGames': 0, 'moves': 30}
r = requests.get(f'https://explorer.lichess.ovh/master', params = payload)
if r.status_code == 200:
time.sleep(0.2)
r = r.json()
break
if r.status_code == 429:
print("Waiting 5 seconds")
time.sleep(5)
continue
# FENdict[fen] = r
matches = r['white'] + r['black'] + r['draws']
return matches
fen = 'rn1qkb1r/pp3ppp/2p1pn2/3p1b2/2PP4/1Q3NP1/PP2PPBP/RNB1K2R b KQkq -'
check_fen_against_mastersdb(fen)
# Identify when a Lichess game was last in "theory" (aka matched at least one masters DB game)
def find_last_ply_found_in_mastersdb(id, start_ply = 0):
pgn = client.games.export(id, as_pgn=True)
game_num = 0
while True:
with io.StringIO(pgn) as f:
game = chess.pgn.read_game(f)
if not game or game_num >= 1:
break
game_num += 1
board = game.board()
masters_matches = 99999
while masters_matches > 0:
for n, move in enumerate(game.mainline_moves()):
if n < start_ply:
board.push(move)
else:
masters_matches = check_fen_against_mastersdb(board.epd())
if masters_matches == 0: # stop when no matches are found
break
else:
board.push(move)
return n-1
find_last_ply_found_in_mastersdb('5fULpU0u')
# Check a ply in a game against the masters DB and return the number of matches
def check_ply_against_mastersdb(id, ply):
pgn = client.games.export(id, as_pgn=True)
game_num = 0
while True:
with io.StringIO(pgn) as f:
game = chess.pgn.read_game(f)
if not game or game_num >= 1:
break
game_num += 1
board = game.board()
for n, move in enumerate(game.mainline_moves()):
if n < ply:
board.push(move)
else:
matches = check_fen_against_mastersdb(board.epd())
break
return matches
# Find game with the latest ply matched in the masters DB
def find_latest_deviation_from_mastersdb(ids):
latest_deviation_ply = 9999
latest_deviation_id = ''
for i in range(0, len(ids)):
print('Checking game', i+1, '/', len(ids), ':', ids[i])
if i == 0:
latest_deviation_ply = find_last_ply_found_in_mastersdb(ids[i], 1)
latest_deviation_id = ids[i]
print('Current latest ply matched in masters DB:', latest_deviation_ply, 'in', latest_deviation_id)
if i > 0:
new_matches = check_ply_against_mastersdb(ids[i], latest_deviation_ply)
if new_matches == 0:
continue
if new_matches > 0:
latest_deviation_ply = find_last_ply_found_in_mastersdb(ids[i], latest_deviation_ply)
latest_deviation_id = ids[i]
print('New latest ply matched in masters DB:', latest_deviation_ply, 'in', latest_deviation_id)
if i > 0 and i % 100 == 0:
print('Current latest ply matched in masters DB:', latest_deviation_ply, 'in', latest_deviation_id)
return [latest_deviation_id, latest_deviation_ply]
find_latest_deviation_from_mastersdb(['0oht8pGP', 'JSyE55lM', '42UIwlRN', 'YsMWRKcn', '2QnQM1ns'])
# More games to test...
# 'rpT0BJkr', 'elDXjTwi', 'HqYHBLxl', 'LmSs5zO5', 'zssiRcV2',
# 'zssiRcV2', 'QqY5HUCU', 'BSB32D9z', '5lFmsjoC', 'SG4cXHBZ',
# '45KDSsll'
###Output
Checking game 1 / 5 : 0oht8pGP
Current latest ply matched in masters DB: 16 in 0oht8pGP
Checking game 2 / 5 : JSyE55lM
Checking game 3 / 5 : 42UIwlRN
New latest ply matched in masters DB: 17 in 42UIwlRN
Checking game 4 / 5 : YsMWRKcn
New latest ply matched in masters DB: 17 in YsMWRKcn
Checking game 5 / 5 : 2QnQM1ns
New latest ply matched in masters DB: 20 in 2QnQM1ns
|
community_tutorials_and_guides/taxi/NYCTaxi-E2E-pandas.ipynb | ###Markdown
Predicting NYC Taxi Fares with RAPIDS RAPIDS is a suite of GPU accelerated data science libraries with APIs that should be familiar to users of Pandas, Dask, and Scikitlearn.This notebook focuses on showing how to use cuDF with Dask & XGBoost to scale GPU DataFrame ETL-style operations & model training out to multiple GPUs on mutliple nodes as part of Google Cloud Dataproc.Anaconda has graciously made some of the NYC Taxi dataset available in a public Google Cloud Storage bucket. We'll use our Dataproc Cluster of GPUs to process it and train a model that predicts the fare amount.For EDA we show the examples using [Holoviews](http://holoviews.org/) and [hvplot](https://hvplot.holoviz.org/). Best way to install Holoviews is to from `conda-forge` channel `conda install -c conda-forge holoviews`and for hvplot `pyviz` channel. `conda install -c pyviz hvplot`
###Code
%matplotlib inline
import matplotlib.pyplot as plt
import socket, time
import pandas as pd
import xgboost as xgb
#To install Holoviews and hvplot
#conda install -c conda-forge holoviews
#conda install -c pyviz hvplot
import holoviews as hv
from holoviews import opts
import numpy as np
import hvplot.pandas
hv.extension('bokeh')
###Output
_____no_output_____
###Markdown
Inspecting the DataNow that we have a cluster of GPU workers, we'll use [dask-cudf](https://github.com/rapidsai/dask-cudf/) to load and parse a bunch of CSV files into a distributed DataFrame.
###Code
'''if you get 'ModuleNotFoundError: No module named 'gcsfs', run `!pip install gcsfs`
'''
base_path = '/localdisk/benchmark_datasets/yellow-taxi-dataset/'
df_2014 = pd.read_csv(base_path+'2014/yellow_tripdata_2014-01.csv', parse_dates=[' pickup_datetime', ' dropoff_datetime'])
df_2014.head()
###Output
_____no_output_____
###Markdown
Data CleanupAs usual, the data needs to be massaged a bit before we can start adding features that are useful to an ML model.For example, in the 2014 taxi CSV files, there are `pickup_datetime` and `dropoff_datetime` columns. The 2015 CSVs have `tpep_pickup_datetime` and `tpep_dropoff_datetime`, which are the same columns. One year has `rate_code`, and another `RateCodeID`.Also, some CSV files have column names with extraneous spaces in them.Worst of all, starting in the July 2016 CSVs, pickup & dropoff latitude and longitude data were replaced by location IDs, making the second half of the year useless to us.We'll do a little string manipulation, column renaming, and concatenating of DataFrames to sidestep the problems.
###Code
#Dictionary of required columns and their datatypes
must_haves = {
'pickup_datetime': 'datetime64[s]',
'dropoff_datetime': 'datetime64[s]',
'passenger_count': 'int32',
'trip_distance': 'float32',
'pickup_longitude': 'float32',
'pickup_latitude': 'float32',
'rate_code': 'int32',
'dropoff_longitude': 'float32',
'dropoff_latitude': 'float32',
'fare_amount': 'float32'
}
def clean(ddf, must_haves):
# replace the extraneous spaces in column names and lower the font type
tmp = {col:col.strip().lower() for col in list(ddf.columns)}
ddf = ddf.rename(columns=tmp)
ddf = ddf.rename(columns={
'tpep_pickup_datetime': 'pickup_datetime',
'tpep_dropoff_datetime': 'dropoff_datetime',
'ratecodeid': 'rate_code'
})
# ddf['pickup_datetime'] = ddf['pickup_datetime'].astype('datetime64[ms]')
# ddf['dropoff_datetime'] = ddf['dropoff_datetime'].astype('datetime64[ms]')
for col in ddf.columns:
if col not in must_haves:
ddf = ddf.drop(columns=col)
continue
# if column was read as a string, recast as float
if ddf[col].dtype == 'object':
ddf[col] = ddf[col].fillna('-1')
# ddf[col] = ddf[col].astype('float32')
# else:
# downcast from 64bit to 32bit types
# Tesla T4 are faster on 32bit ops
# if 'int' in str(ddf[col].dtype):
# ddf[col] = ddf[col].astype('int32')
# if 'float' in str(ddf[col].dtype):
# ddf[col] = ddf[col].astype('float32')
# ddf[col] = ddf[col].fillna(-1)
return ddf
###Output
_____no_output_____
###Markdown
NOTE: We will realize that some of 2015 data has column name as `RateCodeID` and others have `RatecodeID`. When we rename the columns in the clean function, it internally doesn't pass meta while calling map_partitions(). This leads to the error of column name mismatch in the returned data. For this reason, we will call the clean function with map_partition and pass the meta to it. Here is the link to the bug created for that: https://github.com/rapidsai/cudf/issues/5413
###Code
df_2014 = clean(df_2014, must_haves)
###Output
_____no_output_____
###Markdown
We still have 2015 and the first half of 2016's data to read and clean. Let's increase our dataset.
###Code
df_2015 = pd.read_csv(base_path+'2015/yellow_tripdata_2015-01.csv', parse_dates=['tpep_pickup_datetime', 'tpep_dropoff_datetime'])
df_2015 = clean(df_2015, must_haves)
###Output
_____no_output_____
###Markdown
Handling 2016's Mid-Year Schema Change In 2016, only January - June CSVs have the columns we need. If we try to read base_path+2016/yellow_*.csv, Dask will not appreciate having differing schemas in the same DataFrame.Instead, we'll need to create a list of the valid months and read them independently.
###Code
months = [str(x).rjust(2, '0') for x in range(1, 7)]
valid_files = [base_path+'2016/yellow_tripdata_2016-'+month+'.csv' for month in months]
#read & clean 2016 data and concat all DFs
df_2016 = clean(pd.read_csv(valid_files[0], parse_dates=['tpep_pickup_datetime', 'tpep_dropoff_datetime']), must_haves)
#concatenate multiple DataFrames into one bigger one
taxi_df = pd.concat([df_2014, df_2015, df_2016])
# taxi_df = taxi_df.persist()
taxi_df.dtypes
len(taxi_df)
###Output
_____no_output_____
###Markdown
Exploratory Data Analysis (EDA) Here, we are checking out if there are any non-sensical records and outliers, and in such case, we need to remove them from the dataset.
###Code
# check out if there is any negative total trip time
taxi_df[taxi_df.dropoff_datetime <= taxi_df.pickup_datetime].head()
# check out if there is any abnormal data where trip distance is short, but the fare is very high.
taxi_df[(taxi_df.trip_distance < 10) & (taxi_df.fare_amount > 300)].head()
# check out if there is any abnormal data where trip distance is long, but the fare is very low.
taxi_df[(taxi_df.trip_distance > 50) & (taxi_df.fare_amount < 50)].head()
#Using only 2016-01 data for visuals.
#taxi_df_cdf = clean(cudf.read_csv(valid_files[0]),must_haves)
#Using entire 2016 data for visualization
taxi_df_cdf = taxi_df
###Output
_____no_output_____
###Markdown
The plot below visualizes the histogram of trip_distance and we can see some abnormal trip_distance values for some records. Taking this and also the NYC map coordinates into consideration, we will only select records where tripdistance < 500 miles.
###Code
%%time
#Histogram using cupy and Holoviews
# frequencies, edges = cupy.histogram(x=cupy.array(taxi_df_cdf["trip_distance"]) , bins=20)
# hist = hv.Histogram((np.array(edges.tolist()), np.array(frequencies.tolist())))
#Histogram using hvplot
hist = taxi_df_cdf.hvplot.hist("trip_distance", bins=20, bin_range=(0, 10))
#Customizing the plot
hist.opts(xlabel="trip distance (miles)",ylabel="count",color="green",width=900, height=400)
###Output
_____no_output_____
###Markdown
Similarly, the plot below visualizes the histogram of fare_amount and we can see some abnormal fare_amount values for some records. Taking this and also the NYC map coordinates into consideration, we will only select records where fare_amount < 500$.
###Code
%%time
#Histogram using cupy and Holoviews
# frequencies, edges = cupy.histogram(x=cupy.array(taxi_df_cdf["fare_amount"]) , bins=20)
# hist = hv.Histogram((np.array(edges.tolist()), np.array(frequencies.tolist())))
#Histogram using hvplot
hist = taxi_df_cdf.hvplot.hist("fare_amount", bins=20, bin_range=(0, 50))
#Customizing the plot
hist.opts(xlabel="fare amount ($)",ylabel="count",color="green",width=900, height=400)
%%time
# Plot the number of passengers per trip. We'll remove the records where passenger_count > 5.
# Plotting using Holoviews
#bar = hv.Bars(taxi_df_cdf.groupby("passenger_count").size().to_frame().rename(columns={0:"count"}))
# Plotting using hvplot
df_bar = taxi_df_cdf.groupby("passenger_count").size().to_frame().rename(columns={0:"count"}).reset_index()
bar = df_bar.hvplot.bar(x="passenger_count",y="count")
#Customizing the plot
bar.opts(color="green",width=900, height=400)
###Output
_____no_output_____
###Markdown
EDA visuals and additional analysis yield the filter logic below.
###Code
taxi_df
# apply a list of filter conditions to throw out records with missing or outlier values
query_frags = [
'fare_amount > 1 and fare_amount < 500',
'passenger_count > 0 and passenger_count < 6',
'pickup_longitude > -75 and pickup_longitude < -73',
'dropoff_longitude > -75 and dropoff_longitude < -73',
'pickup_latitude > 40 and pickup_latitude < 42',
'dropoff_latitude > 40 and dropoff_latitude < 42',
'trip_distance > 0 and trip_distance < 500',
'not (trip_distance > 50 and fare_amount < 50)',
'not (trip_distance < 10 and fare_amount > 300)',
'not dropoff_datetime <= pickup_datetime'
]
taxi_df = taxi_df.query(' and '.join(query_frags))
# reset_index and drop index column
taxi_df = taxi_df.reset_index(drop=True)
taxi_df
###Output
_____no_output_____
###Markdown
Adding Interesting FeaturesDask & cuDF provide standard DataFrame operations, but also let you run "user defined functions" on the underlying data. Here we use [dask.dataframe's map_partitions](https://docs.dask.org/en/latest/dataframe-api.htmldask.dataframe.DataFrame.map_partitions) to apply user defined python function on each DataFrame partition.We'll use a Haversine Distance calculation to find total trip distance, and extract additional useful variables from the datetime fields.
###Code
## add features
taxi_df['hour'] = taxi_df['pickup_datetime'].dt.hour
taxi_df['year'] = taxi_df['pickup_datetime'].dt.year
taxi_df['month'] = taxi_df['pickup_datetime'].dt.month
taxi_df['day'] = taxi_df['pickup_datetime'].dt.day
taxi_df['day_of_week'] = taxi_df['pickup_datetime'].dt.weekday
taxi_df['is_weekend'] = (taxi_df['day_of_week']>=5).astype('int32')
#calculate the time difference between dropoff and pickup.
taxi_df['diff'] = taxi_df['dropoff_datetime'].astype('int64') - taxi_df['pickup_datetime'].astype('int64')
taxi_df['diff']=(taxi_df['diff']/1000).astype('int64')
taxi_df['pickup_latitude_r'] = taxi_df['pickup_latitude']//.01*.01
taxi_df['pickup_longitude_r'] = taxi_df['pickup_longitude']//.01*.01
taxi_df['dropoff_latitude_r'] = taxi_df['dropoff_latitude']//.01*.01
taxi_df['dropoff_longitude_r'] = taxi_df['dropoff_longitude']//.01*.01
taxi_df = taxi_df.drop('pickup_datetime', axis=1)
taxi_df = taxi_df.drop('dropoff_datetime', axis=1)
geo_columns = ['pickup_longitude', 'pickup_latitude', 'dropoff_longitude', 'dropoff_latitude']
radians = {x: np.radians(taxi_df[x]) for x in geo_columns}
dlon = radians['pickup_longitude'] - radians['dropoff_longitude']
dlat = radians['pickup_latitude'] - radians['dropoff_latitude']
taxi_df['h_distance'] = 6367 * 2 * np.arcsin(
np.sqrt(
np.sin(dlat / 2)**2
+ np.cos(radians['pickup_latitude']) * np.cos(radians['dropoff_latitude']) * np.sin(dlon / 2)**2))
taxi_df
###Output
_____no_output_____
###Markdown
Pick a Training SetLet's imagine you're making a trip to New York on the 25th and want to build a model to predict what fare prices will be like the last few days of the month based on the first part of the month. We'll use a query expression to identify the `day` of the month to use to divide the data into train and test sets.The wall-time below represents how long it takes your GPU cluster to load data from the Google Cloud Storage bucket and the ETL portion of the workflow.
###Code
#since we calculated the h_distance let's drop the trip_distance column, and then do model training with XGB.
taxi_df = taxi_df.drop('trip_distance', axis=1)
# this is the original data partition for train and test sets.
X_train = taxi_df.query('day < 25')
# create a Y_train ddf with just the target variable
Y_train = X_train[['fare_amount']]
# drop the target variable from the training ddf
X_train = X_train[X_train.columns.difference(['fare_amount'])]
###Output
_____no_output_____
###Markdown
Train the XGBoost Regression ModelThe wall time output below indicates how long it took your GPU cluster to train an XGBoost model over the training set.
###Code
X_train
Y_train
dtrain = xgb.DMatrix(X_train, Y_train)
%%time
trained_model = xgb.train({
'learning_rate': 0.3,
'max_depth': 8,
'objective': 'reg:squarederror',
'subsample': 0.6,
'gamma': 1,
'silent': True,
'verbose_eval': True,
'tree_method':'hist'
},
dtrain,
num_boost_round=100, evals=[(dtrain, 'train')])
ax = xgb.plot_importance(trained_model, height=0.8, max_num_features=10, importance_type="gain")
ax.grid(False, axis="y")
ax.set_title('Estimated feature importance')
ax.set_xlabel('importance')
plt.show()
###Output
_____no_output_____
###Markdown
How Good is Our Model?Now that we have a trained model, we need to test it with the 25% of records we held out.Based on the filtering conditions applied to this dataset, many of the DataFrame partitions will wind up having 0 rows. This is a problem for XGBoost which doesn't know what to do with 0 length arrays. We'll repartition the data.
###Code
X_test = taxi_df.query('day >= 25')
# Create Y_test with just the fare amount
Y_test = X_test[['fare_amount']]
# Drop the fare amount from X_test
X_test = X_test[X_test.columns.difference(['fare_amount'])]
# display test set size
len(X_test)
###Output
_____no_output_____
###Markdown
Calculate Prediction
###Code
# generate predictions on the test set
'''feed X_test as a dask.dataframe'''
booster = trained_model
# history = trained_model['history'] # "History" is a dictionary containing evaluation results
# booster.set_param({'predictor': 'gpu_predictor'})
prediction = pd.Series(booster.predict(xgb.DMatrix(X_test)))
prediction.head()
# prediction = prediction.map_partitions(lambda part: cudf.Series(part)).reset_index(drop=True)
actual = Y_test['fare_amount'].reset_index(drop=True)
###Output
_____no_output_____
###Markdown
NOTE: We mapped each partition of the result from `xgb.dask.predict` into `cudf.Series` to be able to substract it from `actual` data. Here is the issue asking XGBoost to solve that before returning data from `xgb.dask.predict` https://github.com/dmlc/xgboost/issues/5823issuecomment-648526888
###Code
prediction.head()
actual.head()
# Calculate RMSE
squared_error = ((prediction-actual)**2)
# compute the actual RMSE over the full test set
np.sqrt(squared_error.mean())
###Output
_____no_output_____
###Markdown
Save Trained Model for Later Use¶ To make a model maximally useful, you need to be able to save it for later use. We'll use Google Cloud Storage to persist the trained model in a dill file.
###Code
import gcsfs, dill
fs = gcsfs.GCSFileSystem()
# replace with a bucket you own
bucket = 'rapidsai-test-1/'
with fs.open(bucket+'trained_model.dill', 'wb') as file:
dill.dump(trained_model, file)
###Output
_____no_output_____
###Markdown
As an alternative, you can save the trained model on your system as below.
###Code
# Save the model to file
booster.save_model('xgboost-model')
print('Training evaluation history:', history)
###Output
_____no_output_____
###Markdown
Reload a Saved Model from Disk You can also read the saved model back out of Google Cloud Storage and into a normal XGBoost model object.
###Code
with fs.open(bucket+'trained_model.dill', 'rb') as file:
model_from_disk = dill.load(file)
# Generate predictions on the test set again, but this time using the reloaded model
prediction = xgb.dask.predict(client, model_from_disk, X_test).persist()
wait(prediction)
# Verify that the predictions result in the same RMSE error
prediction = prediction.map_partitions(lambda part: cudf.Series(part)).reset_index(drop=True)
actual = Y_test['fare_amount'].reset_index(drop=True)
squared_error = ((prediction-actual)**2)
# compute the actual RMSE over the full test set
cupy.sqrt(squared_error.mean().compute())
###Output
_____no_output_____ |
examples/measuring/do2d_multi.ipynb | ###Markdown
Buffered Messurenment SR830Example notebook of buffered measurement for a bundle of SR830 lock-ins. This notebook is shipped together with the file do2d_multi.py containing the function do2d_multi. The do2d_multi is just a function wrapping the QCoDeS Measurement context manager. The do2d_multi takes a list of SR830 lock-ins and perform a buffered measurement, of either channel 1, channel 2 or both. Optionally a list of non-buffered parameters can be provided to be measured on the same grid as the lock-ins. However, adding a nonbuffered parameter will slow the measurement down.
###Code
# IMPORTS
import qcodes as qc
import os
import numpy as np
from qcodes.instrument_drivers.stanford_research.SR830 import SR830
from qdev_wrappers.measurement_helpers.do2d_multi import do2d_multi
from qcodes.instrument.base import Instrument
from qcodes.utils.validators import Numbers, Arrays
import qcodes.instrument_drivers.agilent.Agilent_34400A as agi
from qcodes.dataset.plotting import plot_dataset
###Output
_____no_output_____
###Markdown
Dummy GeneratorVirtual instrument to be used instead of af DAC or other external setting parameter
###Code
class DummyGenerator(Instrument):
def __init__(self, name, **kwargs):
super().__init__(name, **kwargs)
self.add_parameter('v_start',
initial_value=0,
unit='V',
label='v start',
vals=Numbers(0,1e3),
get_cmd=None,
set_cmd=None)
self.add_parameter('v_stop',
initial_value=1,
unit='V',
label='v stop',
vals=Numbers(1,1e3),
get_cmd=None,
set_cmd=None)
self.add_parameter('v_now',
initial_value=0,
unit='V',
label='v_now',
vals=Numbers(self.v_start(),self.v_stop()),
get_cmd=None,
set_cmd=None)
# The parameter to be set in the outer loop
slow = DummyGenerator('slow')
# The parameter to be set in the inner loop
fast = DummyGenerator('fast')
###Output
_____no_output_____
###Markdown
Connect to and Initialze the SR830s
###Code
sr = SR830('lockin', 'GPIB0::2::INSTR')
sr2 = SR830('lockin2', 'GPIB0::1::INSTR')
a1 = agi.Agilent_34400A('Agilent1', 'GPIB0::4::INSTR')
sr.ch1_display('X')
sr.ch1_ratio('none')
sr2.ch1_display('X')
sr2.ch1_ratio('none')
sr.ch2_display('Y')
sr.ch2_ratio('none')
sr2.ch2_display('Y')
sr2.ch2_ratio('none')
a1.reset()
a1.NPLC.set(10)
###Output
Connected to: Stanford_Research_Systems SR830 (serial:s/n70597, firmware:ver1.07) in 1.02s
Connected to: Stanford_Research_Systems SR830 (serial:s/n47762, firmware:ver1.07) in 0.13s
Connected to: HEWLETT-PACKARD 34401A (serial:0, firmware:10-5-2) in 0.11s
###Markdown
Tuple of lock-ins
###Code
lockins = (sr,sr2)
###Output
_____no_output_____
###Markdown
do2d_multiThis is a do2d to be used for a collection of SR830. Args:* param_slow: The QCoDeS parameter to sweep over in the outer loop* start_slow: Starting point of sweep in outer loop* stop_slow: End point of sweep in the outer loop* num_points_slow: Number of points to measure in the outer loop* delay_slow: Delay after setting parameter in the outer loop* param_fast: The QCoDeS parameter to sweep over in the inner loop* start_fast: Starting point of sweep in inner loop* stop_fast: End point of sweep in the inner loop* num_points_fast: Number of points to measure in the inner loop* delay_fast: Delay after setting parameter before measurement is performed* lockins: Tuple of lockins* devices_no_buffer: Iterable of Parameters to be measured alongside the lockins* write_period: The time after which the data is actually written to the database.* threading: For each element which are True, write_in_background, buffer_reset, and send_trigger and get_trace will be threaded respectively* channels: channels to get from the buffer. 0 gets both channels* attempts_to_get: Maximum number of attempts to try to get the buffer if it fails* delay_fast_increase: Amount to increase delay_fast if getting the buffer fails
###Code
datatuple = do2d_multi(param_slow = slow.v_now,
start_slow = 0,
stop_slow = 0.5,
num_points_slow = 10,
delay_slow = 0.05,
param_fast = fast.v_now,
start_fast = 0,
stop_fast = 0.5,
num_points_fast = 10,
delay_fast = 0.07,
lockins = lockins,
write_period = 1,
threading=[True,False,False,True],
channels = 0,
attempts_to_get=50,
delay_fast_increase=0.00
)
# ploting the data
plot_dataset(datatuple[0])
###Output
_____no_output_____
###Markdown
Adding a non-buffered parameter to measure
###Code
devices_no_buffer = (a1.volt,)
datatuple = do2d_multi(param_slow = slow.v_now,
start_slow = 0,
stop_slow = 0.5,
num_points_slow = 10,
delay_slow = 0.05,
param_fast = fast.v_now,
start_fast = 0,
stop_fast = 0.5,
num_points_fast = 10,
delay_fast = 0.07,
lockins = lockins,
devices_no_buffer = devices_no_buffer,
channels = 1,
threading=[True,False,False,True],
attempts_to_get=50,
delay_fast_increase=0.00
)
plot_dataset(datatuple[0])
plot_dataset(datatuple[1])
###Output
_____no_output_____ |
prob-stats-data-analysis/foundational/probfunctions-displaying.ipynb | ###Markdown
Probability functions and displaying data How probability is best formalised and represented, mathematically as well as visually, depends on the type of data you have. In the following, we will use $X$ to indicate a random variable taking values in space $\Omega$. To mathematically express probabilities attached to certain values in $\Omega$, people use the *probability mass function* in the case of discrete variables and *probability density functions* in the case of continuous variables. Let's see them both. The Probability Mass FunctionThe *probability mass function*, aka **pmf**, expresses, at each possible value of a discrete random variable, the probability related to it. We'll call it $p_m$. So, the pfm expresses the probability that $X$ takes a given value $x$:$$p_m(x) = P(X=x) \ ,$$and by normalisation property of probability it obeys$$\sum_{x \in \Omega} p_m(x) = 1$$ The Probability Density FunctionThe **pdf**, *probability density function*, which we'll here call $p_d(x)$, expresses the probability for a continuous random variable $X$ to take values within a certain range, so it is effectively a density of probability.What this means is that taken range of values $[x_a, x_b]$, we have $$P(x_a \leq X \leq x_b) = \int_{x_a}^{x_b} \text{d} x \ p_d(x) \ ,$$and by the normalisation property of probability we have$$\int_\Omega \text{d} x \ p_d(x) = 1$$ Some distributions and using histograms Histogramming data means segmenting it into ranges (*bins*) and counting how many data points fall in each range. It is what you typically do when you have some real-world data and you need to understand how it is distributed. A uniform distribution Let's say we uniformly extract $10^5$ integer data points in the interval $[0, 10)$:
###Code
n = 100000
data = [random.randint(0, 9) for i in range(n)]
###Output
_____no_output_____
###Markdown
and then let's compute, for each of the 10 possible values, how many of these points are there, which is equivalent to say that we are binning with a bin width of 1:
###Code
bins = np.arange(0, 11, 1)
# Count the number of items falling in each bin
bin_counts = [data.count(item) for item in range(0, 10)]
plt.bar(range(0, 10), bin_counts, width=1, edgecolor='k')
plt.xticks(bins)
plt.title('Histogram of $10^5$ uniformly distributed data, counts')
plt.xlabel('Value')
plt.ylabel('Count items')
plt.show();
###Output
_____no_output_____
###Markdown
It's plain visible that bins contain pretty much the same number of values, namely around $10^4$, which is our total divided by the number of bins itself. Indeed, the difference between the highest and the lowest counts in a bins is
###Code
max(bin_counts) - min(bin_counts)
###Output
_____no_output_____
###Markdown
corresponding to a proportion of the total number of points equal to
###Code
(max(bin_counts) - min(bin_counts) ) / n
###Output
_____no_output_____
###Markdown
very little! This histogram is doing a good job in showing us the data is uniformly distributed. What we have shown is, again, the *number* of values in each bin, which is not the probability. In order to have a PMF instead, we'd have to ideally take, for each of the values extracted, its count and then divide it by the total number of values to obtain frequency counts. Note that these are the probabilities of each possible value and they sum up to 1:
###Code
freq_counts = [item / n for item in bin_counts]
sum(freq_counts)
###Output
_____no_output_____
###Markdown
Then we can easily plot them to obtain the PMF histogram:
###Code
plt.bar(range(0, 10), freq_counts, width=1, edgecolor='k')
plt.xticks(range(0, 10))
plt.title('Histogram of $10^5$ uniformly distributed data, probs')
plt.xlabel('Value')
plt.ylabel('Frequency (prob) items')
plt.show();
###Output
_____no_output_____
###Markdown
A gaussian distribution Now let's consider a gaussian distribution instead, taking the same amount ($10^5$) of numbers and plotting the bins counts again. This time we extract float numbers, randomly sampled from a gaussian distribution of mean 0 and standard deviation 1. We then separate the range in 20 bins and plot the histogram of the counts of each bin as above. We use a line to signify that effectively our variable is meant to be continuous.We attribute counts for a bin to the middle point of the bin.
###Code
data = np.random.normal(size=n)
bins = 20 # choose to separate into 20 bins
hist = np.histogram(data, bins=bins)
hist_vals, bin_edges = hist[0], hist[1]
bin_mids = [(bin_edges[i] + bin_edges[i+1])/2 for i in range(len(bin_edges) -1)] # mids of bins again
plt.plot(bin_mids, hist_vals, marker='o')
plt.title('Histogram $10^5$ normally distributed data')
plt.xlabel('Bin mid')
plt.ylabel('Count items')
plt.show();
###Output
_____no_output_____
###Markdown
Each bin is large
###Code
bin_edges[1] - bin_edges[0]
###Output
_____no_output_____
###Markdown
So, it is quite clear from the plot that the mean is indeed at 0. We can also do the same histogram but showing the pdf instead:
###Code
bins = 20
hist = np.histogram(data, bins=bins, density=True)
hist_vals, bin_edges = hist[0], hist[1]
bin_mids = [(bin_edges[i] + bin_edges[i+1])/2 for i in range(len(bin_edges) -1)]
plt.plot(bin_mids, hist_vals, marker='o')
plt.title('Histogram $10^5$ normally distributed data')
plt.xlabel('Bin mid')
plt.ylabel('Count items')
plt.show();
###Output
_____no_output_____
###Markdown
Because what we plotted above here^ is a density of probability, what sums up to 1 is not those values but the product of value times the bin width:
###Code
sum([(bin_edges[i+1] - bin_edges[i]) * hist_vals[i] for i in range(len(hist_vals))])
###Output
_____no_output_____
###Markdown
Effectively indeed, if we take for instance the first bin, its density represents the probability of being in that bin divided by the bin width itself, which is:
###Code
hist_vals[0]
###Output
_____no_output_____
###Markdown
Using boxplots as a visualisation tool Another very useful and quite comprehensive way to display distributions is through the use of *boxplots*. Boxplots let you see, in one go, the quartiles, the mean and the potential outliers in a distribution.To start off with, let's generate $10^5$ random numbers extracting them from a (in order):* uniform distribution* gaussian distribution with mean 0 and standard deviation 1* exponential distribution* Zipf distribution $\propto x^{-2}$ Let's set the styles of our boxplots: the mean will be a red romboid point, the median a red line, the outliers blue circles.Also, we identify as outliers those points which exceed the first or the third quartiles (on the respective sides) by 1.5 their value, according to [Tukey's criterion](http://www.physics.csbsju.edu/stats/box2.html).Following on, we create boxplots for each of our distributions.
###Code
# Extract the random points
n = 100000 # the number of points to extract
u = np.random.uniform(size=n)
g = np.random.normal(size=n) # gaussian
e = np.random.exponential(size=n) # exponential
z = np.random.zipf(a=2, size=n) # Zipf (power-law) x^{-2}
# Set style of the plot
# Style of the point for the mean
meanpointprops = dict(marker='D', markeredgecolor='black', markerfacecolor='firebrick')
# Style of the outliers points
flierpointprops = dict(marker='o', markeredgecolor='blue', linestyle='none')
# Style of the median line
medianlineprops = dict(linewidth=2.5, color='firebrick')
# Uniform distribution
ax = plt.subplot()
ax.boxplot(u, whis=1.5, showmeans=True, vert=False,
meanprops=meanpointprops, flierprops=flierpointprops, medianprops=medianlineprops)
plt.title('Box plot of a uniform distribution')
plt.show();
###Output
_____no_output_____
###Markdown
Clearly the mean is in the middle of values and there are no outliers.
###Code
# The gaussian distribution
ax = plt.subplot()
ax.boxplot(g, whis=1.5, showmeans=True, vert=False,
meanprops=meanpointprops, flierprops=flierpointprops, medianprops=medianlineprops)
plt.title('Box plot of a gaussian distribution with mean 0 and std 1')
plt.show();
###Output
_____no_output_____
###Markdown
The symmetrical tail of "outliers" is visibile.
###Code
# The exponential distribution
ax = plt.subplot()
ax.boxplot(e, whis=1.5, showmeans=True, vert=False,
meanprops=meanpointprops, flierprops=flierpointprops, medianprops=medianlineprops)
plt.title('Box plot of an exponential distribution')
plt.show();
###Output
_____no_output_____
###Markdown
This time the distribution is heavily non-symmetrical and the tail of exceeding values is clear.
###Code
# The Zipf distribution
ax = plt.subplot()
ax.boxplot(z, whis=1.5, showmeans=True, vert=False,
meanprops=meanpointprops, flierprops=flierpointprops, medianprops=medianlineprops)
plt.title('Box plot of a Zipf distribution')
plt.show();
###Output
_____no_output_____ |
data/Output-Python/852021_FDefaultCapsol.ipynb | ###Markdown
Tirmzi Analysisn=1000 m+=1000 nm-=120 istep= 4 min=150 max=700
###Code
import sys
sys.path
import matplotlib.pyplot as plt
import numpy as np
import os
from scipy import signal
ls
import capsol.newanalyzecapsol as ac
ac.get_gridparameters
import glob
folders = glob.glob("Fortran/*/")
folders
all_data= dict()
for folder in folders:
params = ac.get_gridparameters(folder + 'capsol.in')
data = ac.np.loadtxt(folder + 'Z-U.dat')
process_data = ac.process_data(params, data, smoothing=False, std=5*10**-9)
all_data[folder]= (process_data)
all_params= dict()
for folder in folders:
params=ac.get_gridparameters(folder + 'capsol.in')
all_params[folder]= (params)
all_data
all_data.keys()
for key in {key: params for key, params in all_params.items() if params['Thickness_sample'] == .5}:
data=all_data[key]
thickness =all_params[key]['Thickness_sample']
rtip= all_params[key]['Rtip']
er=all_params[key]['eps_r']
plt.plot(data['z'], data['c'], label= f'{rtip} nm, {er}, {thickness} nm')
plt.title('C v. Z for 1nm thick sample')
plt.ylabel("C(m)")
plt.xlabel("Z(m)")
plt.legend()
plt.savefig("C' v. Z for 1nm thick sample 06-28-2021.png")
###Output
_____no_output_____
###Markdown
cut off last experiment because capacitance was off the scale
###Code
for key in {key: params for key, params in all_params.items() if params['Thickness_sample'] == .5}:
data=all_data[key]
thickness=all_params[key]['Thickness_sample']
rtip= all_params[key]['Rtip']
er=all_params[key]['eps_r']
s=slice(4,-3)
plt.plot(data['z'][s], data['cz'][s], label=f'{rtip} nm, {er}, {thickness} nm' )
plt.title('Cz vs. Z for 1.0nm')
plt.ylabel("Cz")
plt.xlabel("Z(m)")
plt.legend()
plt.savefig("Cz v. Z for varying sample thickness, 06-28-2021.png")
for key in {key: params for key, params in all_params.items() if params['Thickness_sample'] == .5}:
data=all_data[key]
thickness=all_params[key]['Thickness_sample']
rtip= all_params[key]['Rtip']
er=all_params[key]['eps_r']
s=slice(5,-5)
plt.plot(data['z'][s], data['czz'][s], label=f'{rtip} nm, {er}, {thickness} nm' )
plt.title('Czz vs. Z for 1.0nm')
plt.ylabel("Czz")
plt.xlabel("Z(m)")
plt.legend()
plt.savefig("Czz v. Z for varying sample thickness, 06-28-2021.png")
params
for key in {key: params for key, params in all_params.items() if params['Thickness_sample'] == .5}:
data=all_data[key]
thickness=all_params[key]['Thickness_sample']
rtip= all_params[key]['Rtip']
er=all_params[key]['eps_r']
s=slice(8,-8)
plt.plot(data['z'][s], data['alpha'][s], label=f'{rtip} nm, {er}, {thickness} nm' )
plt.title('alpha vs. Z for 1.0nm')
plt.ylabel("$\\alpha$")
plt.xlabel("Z(m)")
plt.legend()
plt.savefig("Alpha v. Z for varying sample thickness, 06-28-2021.png")
data
from scipy.optimize import curve_fit
def Cz_model(z, a, n, b,):
return(a*z**n + b)
all_data.keys()
data= all_data['capsol-calc\\0001-capsol\\']
z= data['z'][1:-1]
cz= data['cz'][1:-1]
popt, pcov= curve_fit(Cz_model, z, cz, p0=[cz[0]*z[0], -1, 0])
a=popt[0]
n=popt[1]
b=popt[2]
std_devs= np.sqrt(pcov.diagonal())
sigma_a = std_devs[0]
sigma_n = std_devs[1]
model_output= Cz_model(z, a, n, b)
rmse= np.sqrt(np.mean((cz - model_output)**2))
f"a= {a} ± {sigma_a}"
f"n= {n}± {sigma_n}"
model_output
"Root Mean Square Error"
rmse/np.mean(-cz)
###Output
_____no_output_____ |
preprocessing_pipeline/eda.ipynb | ###Markdown
This notebook will focus on performing EDA on the sales data
###Code
__author__ = "konwar.m"
__copyright__ = "Copyright 2022, AI R&D"
__credits__ = ["konwar.m"]
__license__ = "Individual Ownership"
__version__ = "1.0.1"
__maintainer__ = "konwar.m"
__email__ = "[email protected]"
__status__ = "Development"
###Output
_____no_output_____
###Markdown
The following steps would be covered so as to perform EDA on sales data 1. Data Sourcing2. Data Cleaning3. Univariate Analysis4. Bivariate Analysis5. Multivariate Analysis 1. Data Sourcing: This dataset is collected from a public data source platform known as Kaggle. Please follow this link to get this: https://www.kaggle.com/c/competitive-data-science-predict-future-sales/dataAlso find crucial information regarding this dataset:**File descriptions** sales_train.csv - the training set. Daily historical data from January 2013 to October 2015. test.csv - the test set. You need to forecast the sales for these shops and products for November 2015. sample_submission.csv - a sample submission file in the correct format. items.csv - supplemental information about the items/products. item_categories.csv - supplemental information about the items categories. shops.csv- supplemental information about the shops.**Data fields** ID - an Id that represents a (Shop, Item) tuple within the test set shop_id - unique identifier of a shop item_id - unique identifier of a product item_category_id - unique identifier of item category item_cnt_day - number of products sold. You are predicting a monthly amount of this measure item_price - current price of an item date - date in format dd/mm/yyyy date_block_num - a consecutive month number, used for convenience. January 2013 is 0, February 2013 is 1,..., October 2015 is 33 item_name - name of item shop_name - name of shop item_category_name - name of item category 2. Data CleaningIrregularities in a dataset are of different types. Please refer to details below Missing Values Duplicate ValuesIncorrect Format Incorrect Headers Anomalies/Outliers
###Code
# Importing useful libraries.
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
%matplotlib inline
os.chdir('..')
os.getcwd()
# Read the data set of "Marketing Analysis" in data.
retail_data = pd.read_csv(os.path.join('datasets', 'sales_train.csv'))
# Printing the data
retail_data.head()
###Output
_____no_output_____
###Markdown
Check for missing values There are mainly three types of missing values. **MCAR(Missing completely at random):** These values do not depend on any other features. **MAR(Missing at random):** These values may be dependent on some other features. **MNAR(Missing not at random):** These missing values have some reason for why they are missing.
###Code
# Checking the missing values
retail_data.isnull().sum()
###Output
_____no_output_____
###Markdown
If there are lesser number of missing values in any of the column, they can be easily removed from the dataset but the best way to handle such situation is to replace those missing values with 'mean' or 'median' in case of numerical values or 'mode' in case of categorical column Check for duplicates records
###Code
duplicate = retail_data.duplicated()
print(duplicate.sum())
retail_data[duplicate]
###Output
6
###Markdown
Since, there are only 6 duplicates, we will remove them from the retail dataset and re-evaluate duplications again
###Code
retail_data.drop_duplicates(inplace=True)
duplicate = retail_data.duplicated()
print(duplicate.sum())
###Output
0
###Markdown
Handling outliersThere are two types of outliers: **1. Univariate outliers:** Univariate outliers are the data points whose values lie beyond the range of expected values based on one variable. **2. Multivariate outliers:** While plotting data, some values of one variable may not lie beyond the expected range, but when you plot the data with some other variable, these values may lie far from the expected value.Box plot is used normally to check for outliers
###Code
retail_data.boxplot(column=['item_cnt_day'])
###Output
_____no_output_____
###Markdown
3. Univariate Analysis **Categorical Unordered Univariate Analysis:** Its applied on categorical columns once we identify which columns are object types from the dataset
###Code
retail_data.dtypes
# Let's calculate the percentage of unique date counts.
retail_data.head(100).date.value_counts(normalize=True)
#plot the bar graph of percentage dates
plt.figure(figsize=(10,10))
retail_data.head(1000).date.value_counts(normalize=True).plot.barh()
plt.show()
###Output
_____no_output_____
###Markdown
4.Bivariate Analysis: If we analyze data by taking two variables/columns into consideration from a dataset may be numerical columns one on one. a) Numeric Bivariate Analysis is done by using below techniques:1. Scatter Plot2. Pair Plot and3. Correlation Plot
###Code
#plot the scatter plot of balance and salary variable in data
plt.scatter(retail_data.item_price, retail_data.item_cnt_day)
plt.show()
#plot the pair plot of salary, balance and age in data dataframe.
sns.pairplot(data = retail_data, vars=['item_price','item_cnt_day','shop_id', 'item_id'])
plt.show()
plt.figure(figsize=(8,8))
c= retail_data.corr()
sns.heatmap(c, cmap='BrBG', annot=True)
###Output
_____no_output_____
###Markdown
5. Multivariate AnalysisItem Price, Item Count Per day, Item ID is used to pivot the results
###Code
result = pd.pivot_table(data=retail_data.head(1000), index='shop_id', columns='item_id',values='item_cnt_day')
print(result)
plt.figure(figsize=(15,15))
#create heat map of education vs marital vs response_rate
sns.heatmap(result, annot=True, cmap = 'RdYlGn', center=0.117)
plt.show()
###Output
item_id 785 791 804 810 829 832 839 944 965 \
shop_id
25 1.0 1.333333 1.0 1.0 1.0 1.0 1.0 1.0 1.0
59 NaN NaN NaN NaN NaN NaN NaN NaN NaN
item_id 970 ... 5224 5256 5260 5261 5263 5272 5275 \
shop_id ...
25 1.0 ... 1.0 1.333333 1.0 1.2 1.0 1.428571 1.0
59 NaN ... NaN NaN NaN NaN NaN NaN NaN
item_id 5313 5324 22154
shop_id
25 1.0 1.0 NaN
59 NaN NaN 1.0
[2 rows x 380 columns]
|
notebooks/5.Using of RGB strip/1.Light up a RGB light at any position/Light up a RGB light at any position.ipynb | ###Markdown
@Copyright (C): 2010-2019, Shenzhen Yahboom Tech @Author: Malloy.Yuan @Date: 2019-07-17 10:10:02 @LastEditors: Malloy.Yuan @LastEditTime: 2019-09-17 17:54:19 Import Yahboom officially packaged RGB driver libraryCreate and initialize a programmable RGB object
###Code
from RGB_Lib import Programing_RGB
RGB = Programing_RGB()
###Output
_____no_output_____
###Markdown
Set the first RGB light to white., which is the left rear position RGB light of Jetbot The color parameter in the method is R G B
###Code
RGB.Set_An_RGB(0, 0xFF, 0xFF, 0xFF)
###Output
_____no_output_____
###Markdown
Set all RGB to red
###Code
RGB.Set_All_RGB(0xFF, 0x00, 0x00)
###Output
_____no_output_____
###Markdown
Set all RGB to green
###Code
RGB.Set_All_RGB(0x00, 0xFF, 0x00)
###Output
_____no_output_____
###Markdown
Set all RGB to blue
###Code
RGB.Set_All_RGB(0x00, 0x00, 0xFF)
###Output
_____no_output_____
###Markdown
Close all RGB
###Code
RGB.OFF_ALL_RGB()
###Output
_____no_output_____ |
_notebooks/2021-10-07-kagglestudy3.ipynb | ###Markdown
"[SSUDA] 택시 데이터 분석"- author: Seong Yeon Kim - categories: [SSUDA, jupyter, kaggle, datetime, scale, location, Regression] - image: images/211007.png 데이터 불러오기
###Code
!pip install kaggle
from google.colab import files
files.upload()
!mkdir -p ~/.kaggle
!cp kaggle.json ~/.kaggle/
!chmod 600 ~/.kaggle/kaggle.json
!kaggle competitions download -c nyc-taxi-trip-duration
!unzip train.zip
!unzip test.zip
!unzip sample_submission.zip
###Output
Archive: train.zip
replace train.csv? [y]es, [n]o, [A]ll, [N]one, [r]ename: Y
inflating: train.csv
Archive: test.zip
replace test.csv? [y]es, [n]o, [A]ll, [N]one, [r]ename: Y
inflating: test.csv
Archive: sample_submission.zip
replace sample_submission.csv? [y]es, [n]o, [A]ll, [N]one, [r]ename: Y
inflating: sample_submission.csv
###Markdown
압축되어 있는 데이터라서 압축 풀어줍니다.
###Code
%matplotlib inline
import pandas as pd
from datetime import datetime
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression, Ridge,BayesianRidge
from sklearn.cluster import MiniBatchKMeans
from sklearn.metrics import mean_squared_error
from math import radians, cos, sin, asin, sqrt
import seaborn as sns
import matplotlib
import numpy as np
import matplotlib.pyplot as plt
plt.rcParams['figure.figsize'] = [16, 10]
train = pd.read_csv('./train.csv')
test = pd.read_csv('./test.csv')
###Output
_____no_output_____
###Markdown
데이터 탐색
###Code
train.head()
train.describe()
train.info()
test.info()
###Output
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 625134 entries, 0 to 625133
Data columns (total 9 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 id 625134 non-null object
1 vendor_id 625134 non-null int64
2 pickup_datetime 625134 non-null object
3 passenger_count 625134 non-null int64
4 pickup_longitude 625134 non-null float64
5 pickup_latitude 625134 non-null float64
6 dropoff_longitude 625134 non-null float64
7 dropoff_latitude 625134 non-null float64
8 store_and_fwd_flag 625134 non-null object
dtypes: float64(4), int64(2), object(3)
memory usage: 42.9+ MB
###Markdown
dropoff_datetime 변수가 test에는 없습니다. 도착 시간을 맞추는 예제이기 때문에 그렇습니다. 반응변수 관찰
###Code
plt.figure(figsize=(8,6))
plt.scatter(range(train.shape[0]), np.sort(train.trip_duration.values))
plt.xlabel('index', fontsize=12)
plt.ylabel('trip duration', fontsize=12)
plt.show()
###Output
_____no_output_____
###Markdown
반응변수의 이상치가 많아보입니다. 제거하겠습니다.
###Code
m = np.mean(train['trip_duration'])
s = np.std(train['trip_duration'])
train = train[train['trip_duration'] <= m + 2*s]
train = train[train['trip_duration'] >= m - 2*s]
plt.figure(figsize=(8,6))
plt.scatter(range(train.shape[0]), np.sort(train.trip_duration.values))
plt.xlabel('index', fontsize=12)
plt.ylabel('trip duration', fontsize=12)
plt.show()
###Output
_____no_output_____
###Markdown
이상치는 대부분 제거된 것 같습니다. 다만 일부 데이터가 큰 값을 갖는거 같아요.
###Code
plt.hist(train['trip_duration'].values, bins=100)
plt.xlabel('trip_duration')
plt.ylabel('number of train records')
plt.show()
###Output
_____no_output_____
###Markdown
히스토그램으로 확인하니 그렇습니다. 우측 꼬리가 긴 모양으로 로그변환이 필요해보입니다.
###Code
train['log_trip_duration'] = np.log(train['trip_duration'].values + 1)
plt.hist(train['log_trip_duration'].values, bins=100)
plt.xlabel('log(trip_duration)')
plt.ylabel('number of train records')
plt.show()
sns.distplot(train["log_trip_duration"], bins =100)
###Output
_____no_output_____
###Markdown
확실히 그래프 모양이 괜찮아졌습니다. distplot 함수를 통해 그리기도 하였네요 데이터 전처리
###Code
train = train[train['pickup_longitude'] <= -73.75]
train = train[train['pickup_longitude'] >= -74.03]
train = train[train['pickup_latitude'] <= 40.85]
train = train[train['pickup_latitude'] >= 40.63]
train = train[train['dropoff_longitude'] <= -73.75]
train = train[train['dropoff_longitude'] >= -74.03]
train = train[train['dropoff_latitude'] <= 40.85]
train = train[train['dropoff_latitude'] >= 40.63]
###Output
_____no_output_____
###Markdown
뉴욕의 위도는 (-74.03, -73.75) 경도는 (40.63, 40.85)사이 입니다. 이 값을 벗어나는 위도/경도 데이터를 제거하겠습니다.
###Code
train['pickup_datetime'] = pd.to_datetime(train.pickup_datetime)
test['pickup_datetime'] = pd.to_datetime(test.pickup_datetime)
train.loc[:, 'pickup_date'] = train['pickup_datetime'].dt.date
test.loc[:, 'pickup_date'] = test['pickup_datetime'].dt.date
train['dropoff_datetime'] = pd.to_datetime(train.dropoff_datetime) #Not in Test
###Output
_____no_output_____
###Markdown
to_datetime 함수로 datetime 변수로 바궈주었습니다.
###Code
plt.plot(train.groupby('pickup_date').count()[['id']], 'o-', label='train')
plt.plot(test.groupby('pickup_date').count()[['id']], 'o-', label='test')
plt.title('Trips over Time.')
plt.legend(loc=0)
plt.ylabel('Trips')
plt.show()
###Output
_____no_output_____
###Markdown
트레인과 테스트 데이터를 같이 그리니 유사한 측면을 발견하기가 쉬운것 같아요.1월 하순경 이동횟수가 급격하게 감소한것이 관찰됩니다. 또 5월 하순경 감소세가 또 관찰됩니다.계절적으로 추운것도 있겠지만 작성자는 다른 요인이 있지 않을까 생각하네요.
###Code
import warnings
warnings.filterwarnings("ignore")
plot_vendor = train.groupby('vendor_id')['trip_duration'].mean()
plt.subplots(1,1,figsize=(17,10))
plt.ylim(ymin=800)
plt.ylim(ymax=840)
sns.barplot(plot_vendor.index,plot_vendor.values)
plt.title('Time per Vendor')
plt.legend(loc=0)
plt.ylabel('Time in Seconds')
###Output
No handles with labels found to put in legend.
###Markdown
범위를 800~840으로 두어서 그렇지 두 vendor 간 큰 차이를 보이진 않습니다.
###Code
snwflag = train.groupby('store_and_fwd_flag')['trip_duration'].mean()
plt.subplots(1,1,figsize=(17,10))
plt.ylim(ymin=0)
plt.ylim(ymax=1100)
plt.title('Time per store_and_fwd_flag')
plt.legend(loc=0)
plt.ylabel('Time in Seconds')
sns.barplot(snwflag.index,snwflag.values)
###Output
No handles with labels found to put in legend.
###Markdown
공급업체에 보내기 전 기록이 잘 저장되었는지 나타내는 변수로 꽤 많이 차이가 납니다.작성자는 일부 직원이 이동시간을 정확히 기록하지 못해 발생하는 왜곡이라고 말합니다.
###Code
pc = train.groupby('passenger_count')['trip_duration'].mean()
plt.subplots(1,1,figsize=(17,10))
plt.ylim(ymin=0)
plt.ylim(ymax=1100)
plt.title('Time per store_and_fwd_flag')
plt.legend(loc=0)
plt.ylabel('Time in Seconds')
sns.barplot(pc.index,pc.values)
###Output
No handles with labels found to put in legend.
###Markdown
승객 수는 뚜렷한 여행을 주지 못합니다.승객을 아무도 태우지 않았는데 4분정도 이동한 것은 직원의 실수로 보입니다.
###Code
train.groupby('passenger_count').size()
###Output
_____no_output_____
###Markdown
위치 데이터
###Code
city_long_border = (-74.03, -73.75)
city_lat_border = (40.63, 40.85)
fig, ax = plt.subplots(ncols=2, sharex=True, sharey=True)
ax[0].scatter(train['pickup_longitude'].values[:100000], train['pickup_latitude'].values[:100000],
color='blue', s=1, label='train', alpha=0.1)
ax[1].scatter(test['pickup_longitude'].values[:100000], test['pickup_latitude'].values[:100000],
color='green', s=1, label='test', alpha=0.1)
fig.suptitle('Train and test area complete overlap.')
ax[0].legend(loc=0)
ax[0].set_ylabel('latitude')
ax[0].set_xlabel('longitude')
ax[1].set_xlabel('longitude')
ax[1].legend(loc=0)
plt.ylim(city_lat_border)
plt.xlim(city_long_border)
plt.show()
###Output
_____no_output_____
###Markdown
자세한 코드 관찰은 위치 데이터 분석을 할때 다시 확인하겠습니다.train, test 간 위치 데이터가 매우 유사함을 알 수 있습니다.
###Code
def haversine_array(lat1, lng1, lat2, lng2):
lat1, lng1, lat2, lng2 = map(np.radians, (lat1, lng1, lat2, lng2))
AVG_EARTH_RADIUS = 6371 # in km
lat = lat2 - lat1
lng = lng2 - lng1
d = np.sin(lat * 0.5) ** 2 + np.cos(lat1) * np.cos(lat2) * np.sin(lng * 0.5) ** 2
h = 2 * AVG_EARTH_RADIUS * np.arcsin(np.sqrt(d))
return h
def dummy_manhattan_distance(lat1, lng1, lat2, lng2):
a = haversine_array(lat1, lng1, lat1, lng2)
b = haversine_array(lat1, lng1, lat2, lng1)
return a + b
def bearing_array(lat1, lng1, lat2, lng2):
AVG_EARTH_RADIUS = 6371 # in km
lng_delta_rad = np.radians(lng2 - lng1)
lat1, lng1, lat2, lng2 = map(np.radians, (lat1, lng1, lat2, lng2))
y = np.sin(lng_delta_rad) * np.cos(lat2)
x = np.cos(lat1) * np.sin(lat2) - np.sin(lat1) * np.cos(lat2) * np.cos(lng_delta_rad)
return np.degrees(np.arctan2(y, x))
train.loc[:, 'distance_haversine'] = haversine_array(train['pickup_latitude'].values, train['pickup_longitude'].values, train['dropoff_latitude'].values, train['dropoff_longitude'].values)
test.loc[:, 'distance_haversine'] = haversine_array(test['pickup_latitude'].values, test['pickup_longitude'].values, test['dropoff_latitude'].values, test['dropoff_longitude'].values)
train.loc[:, 'distance_dummy_manhattan'] = dummy_manhattan_distance(train['pickup_latitude'].values, train['pickup_longitude'].values, train['dropoff_latitude'].values, train['dropoff_longitude'].values)
test.loc[:, 'distance_dummy_manhattan'] = dummy_manhattan_distance(test['pickup_latitude'].values, test['pickup_longitude'].values, test['dropoff_latitude'].values, test['dropoff_longitude'].values)
train.loc[:, 'direction'] = bearing_array(train['pickup_latitude'].values, train['pickup_longitude'].values, train['dropoff_latitude'].values, train['dropoff_longitude'].values)
test.loc[:, 'direction'] = bearing_array(test['pickup_latitude'].values, test['pickup_longitude'].values, test['dropoff_latitude'].values, test['dropoff_longitude'].values)
###Output
_____no_output_____
###Markdown
위도/경도를 활용하여 다양한 관측값을 나타내는 함수입니다.이해하기에 조금 벅차서 일단 다양한 변수를 추가해줄수 있구나 하고 넘어갔네요.
###Code
coords = np.vstack((train[['pickup_latitude', 'pickup_longitude']].values,
train[['dropoff_latitude', 'dropoff_longitude']].values))
sample_ind = np.random.permutation(len(coords))[:500000]
kmeans = MiniBatchKMeans(n_clusters=100, batch_size=10000).fit(coords[sample_ind])
train.loc[:, 'pickup_cluster'] = kmeans.predict(train[['pickup_latitude', 'pickup_longitude']])
train.loc[:, 'dropoff_cluster'] = kmeans.predict(train[['dropoff_latitude', 'dropoff_longitude']])
test.loc[:, 'pickup_cluster'] = kmeans.predict(test[['pickup_latitude', 'pickup_longitude']])
test.loc[:, 'dropoff_cluster'] = kmeans.predict(test[['dropoff_latitude', 'dropoff_longitude']])
###Output
_____no_output_____
###Markdown
np.vstack는 데이터를 묶어주는 함수입니다.위도, 경도 데이터를 클러스트로 묶어주었습니다.
###Code
fig, ax = plt.subplots(ncols=1, nrows=1)
ax.scatter(train.pickup_longitude.values[:500000], train.pickup_latitude.values[:500000], s=10, lw=0,
c=train.pickup_cluster[:500000].values, cmap='autumn', alpha=0.2)
ax.set_xlim(city_long_border)
ax.set_ylim(city_lat_border)
ax.set_xlabel('Longitude')
ax.set_ylabel('Latitude')
plt.show()
###Output
_____no_output_____
###Markdown
군집화가 잘된 것을 시각적으로 확인하였습니다. 날짜 데이터
###Code
train['Month'] = train['pickup_datetime'].dt.month
test['Month'] = test['pickup_datetime'].dt.month
train['DayofMonth'] = train['pickup_datetime'].dt.day
test['DayofMonth'] = test['pickup_datetime'].dt.day
train['Hour'] = train['pickup_datetime'].dt.hour
test['Hour'] = test['pickup_datetime'].dt.hour
train['dayofweek'] = train['pickup_datetime'].dt.dayofweek
test['dayofweek'] = test['pickup_datetime'].dt.dayofweek
###Output
_____no_output_____
###Markdown
픽업된 시간으로 다양한 파생 날짜/시간 데이터를 생성한 모습입니다. datetime 변수이기에 가능한 모습입니다.여기서 dayofweek 변수는 요일변수로 0을 일요일로 생각하여 6을 토요일까지 쓰는 변수입니다.
###Code
train.loc[:, 'avg_speed_h'] = 1000 * train['distance_haversine'] / train['trip_duration']
train.loc[:, 'avg_speed_m'] = 1000 * train['distance_dummy_manhattan'] / train['trip_duration']
fig, ax = plt.subplots(ncols=3, sharey=True)
ax[0].plot(train.groupby('Hour').mean()['avg_speed_h'], 'bo-', lw=2, alpha=0.7)
ax[1].plot(train.groupby('dayofweek').mean()['avg_speed_h'], 'go-', lw=2, alpha=0.7)
ax[2].plot(train.groupby('Month').mean()['avg_speed_h'], 'ro-', lw=2, alpha=0.7)
ax[0].set_xlabel('Hour of Day')
ax[1].set_xlabel('Day of Week')
ax[2].set_xlabel('Month of Year')
ax[0].set_ylabel('Average Speed')
fig.suptitle('Average Traffic Speed by Date-part')
plt.show()
###Output
_____no_output_____
###Markdown
정확히 이해하진 못했지만 distance_haversine가 위치 변수를 보고 만든 거리 변수입니다.그렇기 때문에 거리 / 시간 = 평균속도 변수를 만들었습니다. 이 평균속도를 시각/요일/달 별로 얼마나 다른지 시각화했습니다.물론 분모인 시간이 반응변수 이기 때문에 분석에 사용할수는 없습니다.보통 오전 5시~9시, 오후 5시(17시) ~ 7시(19시) 사이가 가장 도로가 혼잡해 속도가 떨어집니다.예상과 어느정도 일치하면서도 출/퇴근 이외 근무시간도 속도가 출/퇴근 시간과 비슷하게 떨어집니다.또 금토일의 평균속도가 상대적으로 빠르며 달별로는 겨울의 평균속도가 빠릅니다. 원핫인코딩
###Code
vendor_train = pd.get_dummies(train['vendor_id'], prefix='vi', prefix_sep='_')
vendor_test = pd.get_dummies(test['vendor_id'], prefix='vi', prefix_sep='_')
passenger_count_train = pd.get_dummies(train['passenger_count'], prefix='pc', prefix_sep='_')
passenger_count_test = pd.get_dummies(test['passenger_count'], prefix='pc', prefix_sep='_')
store_and_fwd_flag_train = pd.get_dummies(train['store_and_fwd_flag'], prefix='sf', prefix_sep='_')
store_and_fwd_flag_test = pd.get_dummies(test['store_and_fwd_flag'], prefix='sf', prefix_sep='_')
cluster_pickup_train = pd.get_dummies(train['pickup_cluster'], prefix='p', prefix_sep='_')
cluster_pickup_test = pd.get_dummies(test['pickup_cluster'], prefix='p', prefix_sep='_')
cluster_dropoff_train = pd.get_dummies(train['dropoff_cluster'], prefix='d', prefix_sep='_')
cluster_dropoff_test = pd.get_dummies(test['dropoff_cluster'], prefix='d', prefix_sep='_')
month_train = pd.get_dummies(train['Month'], prefix='m', prefix_sep='_')
month_test = pd.get_dummies(test['Month'], prefix='m', prefix_sep='_')
dom_train = pd.get_dummies(train['DayofMonth'], prefix='dom', prefix_sep='_')
dom_test = pd.get_dummies(test['DayofMonth'], prefix='dom', prefix_sep='_')
hour_train = pd.get_dummies(train['Hour'], prefix='h', prefix_sep='_')
hour_test = pd.get_dummies(test['Hour'], prefix='h', prefix_sep='_')
dow_train = pd.get_dummies(train['dayofweek'], prefix='dow', prefix_sep='_')
dow_test = pd.get_dummies(test['dayofweek'], prefix='dow', prefix_sep='_')
###Output
_____no_output_____
###Markdown
범주형 변수들을 전부 원핫인코딩을 했습니다.prefix 와 prefix_sep 으로 원핫인코딩 변수 이름도 설정할수 있네요.
###Code
passenger_count_test = passenger_count_test.drop('pc_9', axis = 1)
###Output
_____no_output_____
###Markdown
다만 9명이 탑승한 2건은 표본이 너무 적어 과적합될수도 있고 직관적으로도 말이 안되서 열을 삭제합니다.
###Code
train = train.drop(['id','vendor_id','passenger_count','store_and_fwd_flag','Month','DayofMonth','Hour','dayofweek','pickup_datetime',
'pickup_date','pickup_longitude','pickup_latitude','dropoff_longitude','dropoff_latitude'],axis = 1)
Test_id = test['id']
test = test.drop(['id','vendor_id','passenger_count','store_and_fwd_flag','Month','DayofMonth','Hour','dayofweek', 'pickup_datetime',
'pickup_date', 'pickup_longitude','pickup_latitude','dropoff_longitude','dropoff_latitude'], axis = 1)
train = train.drop(['dropoff_datetime','avg_speed_h','avg_speed_m','trip_duration'], axis = 1)
###Output
_____no_output_____
###Markdown
원핫인코딩 된 변수들, 시각화를 위해 만들었던 변수들, 변환한 변수들, id 등 필요없는 변수를 제거합니다.
###Code
Train_Master = pd.concat([train,
vendor_train,
passenger_count_train,
store_and_fwd_flag_train,
cluster_pickup_train,
cluster_dropoff_train,
month_train,
dom_train,
hour_test,
dow_train
], axis=1)
Test_master = pd.concat([test,
vendor_test,
passenger_count_test,
store_and_fwd_flag_test,
cluster_pickup_test,
cluster_dropoff_test,
month_test,
dom_test,
hour_test,
dow_test], axis=1)
Train_Master.shape,Test_master.shape
###Output
_____no_output_____
###Markdown
원핫인코딩했던 변수들을 합쳐줍니다. 모델 적합
###Code
X_train = Train_Master.drop(['log_trip_duration'], axis=1)
Y_train = Train_Master["log_trip_duration"]
Y_train = Y_train.reset_index().drop('index',axis = 1)
###Output
_____no_output_____
###Markdown
이 코드 이후로 모델적합을 해야하는데 코랩에서 계속 램이 부족하다고 하네요.데이터도 크고, 열 개수도 원핫인코딩으로 늘려서 그런거 같습니다.XGB였다가 LGB로 바꾸고, 노말모델로 하고 어떻게 해도 계속 램이 부족해서 실행이 안되네요.코드를 리뷰하는 목적이고 요즘 시간이 넉넉하지 못해서 여기까지 하겠습니다.
###Code
from lightgbm import LGBMRegressor
model = LGBMRegressor()
model.fit(X_train, Y_train)
pred = model.predict(Test_master)
pred = np.exp(pred)
submission = pd.concat([Test_id, pd.DataFrame(pred)], axis=1)
submission.columns = ['id','trip_duration']
submission['trip_duration'] = submission.apply(lambda x : 1 if (x['trip_duration'] <= 0) else x['trip_duration'], axis = 1)
submission.to_csv("./submission.csv", index=False)
!kaggle competitions submit -c nyc-taxi-trip-duration -f submission.csv -m "Message"
###Output
_____no_output_____ |
notebooks/4.a- Convex_optimization_model_with_previous_influence_matrices.ipynb | ###Markdown
=========================================================== Solve the estimation problem with convex optimization model on the supervised dataset from the Jeopardy-like logs ===========================================================Goals:1. Split the data into test and train2. Formulate the convex optimization model3. Compute train and test error Last update: 03 Dec 2019 Imports
###Code
from __future__ import division, print_function, absolute_import, unicode_literals
import cvxpy as cp
import scipy as sp
import pandas as pd
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
import seaborn as sns
from collections import defaultdict
import sys
from sklearn.model_selection import KFold
from sklearn.model_selection import train_test_split
sys.path.insert(0, '../src/')
%matplotlib inline
import utils
from mytimer import Timer
import imp
def reload():
imp.reload(utils)
###Output
_____no_output_____
###Markdown
Parameters
###Code
data_fpath = '/home/omid/Datasets/Jeopardy/supervised_data_with_only_first_influence.pk'
# data_fpath = '/home/omid/Datasets/Jeopardy/supervised_data.pk'
# lambdaa = 1
test_fraction = 0.2
runs = 30
###Output
_____no_output_____
###Markdown
Helper functions
###Code
def compute_matrix_err(true_matrix: np.matrix, pred_matrix: np.matrix, type_str: str = 'corr') -> float:
if type_str == 'frob_norm':
frob_norm_of_difference = np.linalg.norm(true_matrix - pred_matrix)
err = frob_norm_of_difference / np.linalg.norm(true_matrix)
return err
elif type_str == 'corr':
# (r, p) = sp.stats.spearmanr(np.array(true_matrix.flatten())[0], np.array(pred_matrix.flatten())[0])
(r, p) = sp.stats.pearsonr(np.array(true_matrix.flatten())[0], np.array(pred_matrix.flatten())[0])
if p > 0.05:
r = 0
return r
else:
raise ValueError('Wrong type_str was given.')
###Output
_____no_output_____
###Markdown
Loading the data
###Code
data = utils.load_it(data_fpath)
print(len(data['X']))
mats = []
for i in range(len(data['y'])):
mats.append(data['y'][i]['influence_matrix'] / 100)
np.mean(mats, axis=0)
np.std(mats, axis=0)
###Output
_____no_output_____
###Markdown
Formulating the convex optimization problem With only average of previous influence matrices:
###Code
lambdaa = 0.1
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
A1 = element['average_of_previous_influence_matrices']
pred_influence_matrix = A1 * W1 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
constraints += [pred_influence_matrix >= 0]
constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
A1 = element['average_of_previous_influence_matrices']
predicted_influence_matrix = A1 * W1.value + B.value
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
sns.heatmap(W1.value);
sns.heatmap(B.value);
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6257138866867928 +- 0.01604257548685887
uniform: 0.3371604198790437 +- 0.01908804176115327
model: 0.23749912395791864 +- 0.015044850361760172
###Markdown
Logistic function with only first influence matrix
###Code
lambdaa = 0.1
model_errs = []
random_errs = []
uniform_errs = []
for run in range(2):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
A1 = element['first_influence_matrices']
# losses += cp.sum_entries( cp.logistic(-influence_matrix * (A1 * W1 + B)) )
pred_influence_matrix = A1 * W1 + B
losses += cp.sum_entries( cp.kl_div(influence_matrix, pred_influence_matrix) )
# pred_influence_matrix = A1 * W1 + B
# losses += cp.sum_squares(pred_influence_matrix - influence_matrix)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
A1 = element['first_influence_matrices']
predicted_influence_matrix = A1 * W1.value + B.value
# predicted_influence_matrix = utils.make_matrix_row_stochastic(predicted_influence_matrix)
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
sns.heatmap(W1.value)
sns.heatmap(B.value)
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6174122674483592 +- 0.010745327398581805
uniform: 0.32959302075858365 +- 0.0013100443835122044
model: 0.3118306822846595 +- 0.014195220272253922
###Markdown
With only first influence matrix:
###Code
lambdaa = 0.1
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
A1 = element['first_influence_matrices']
pred_influence_matrix = A1 * W1 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
A1 = element['first_influence_matrices']
predicted_influence_matrix = A1 * W1.value + B.value
predicted_influence_matrix = utils.make_matrix_row_stochastic(predicted_influence_matrix)
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
sns.heatmap(W1.value);
sns.heatmap(B.value);
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
# Just the dataset itself:
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6212294304253135 +- 0.01706147914657565
uniform: 0.3308111043908909 +- 0.019409182902447326
model: 0.3044930561618803 +- 0.016874360467564523
###Markdown
With individual performance
###Code
runs = 30
with Timer():
lambdaa = 0.1
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
# A1 = element['individual_performance']
p = element['individual_performance_hardness_weighted']
A1 = np.row_stack([p, p, p, p])
pred_influence_matrix = A1 * W1 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
# A1 = element['individual_performance']
p = element['individual_performance_hardness_weighted']
A1 = np.row_stack([p, p, p, p])
predicted_influence_matrix = A1 * W1.value + B.value
predicted_influence_matrix = utils.make_matrix_row_stochastic(predicted_influence_matrix)
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
sns.heatmap(W1.value);
sns.heatmap(B.value);
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform'])
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6216012908918773 +- 0.01642054499508854
uniform: 0.33076050491720327 +- 0.020687093389048193
model: 0.3277117626599369 +- 0.018084551142832275
###Markdown
With first influence matrix and individual performance (with correlation)
###Code
with Timer():
# injaaaaaaaaaaaaaa
lambdaa = 0.1
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
W2 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
y_train[index]['influence_matrix']))
A1 = element['first_influence_matrices']
p = element['individual_performance_hardness_weighted']
A2 = np.row_stack([p, p, p, p])
pred_influence_matrix = A1 * W1 + A2 * W2 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
y_test[index]['influence_matrix']))
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
# A1 = element['individual_performance']
A1 = element['first_influence_matrices']
p = element['individual_performance_hardness_weighted']
A2 = np.row_stack([p, p, p, p])
predicted_influence_matrix = A1 * W1.value + A2 * W2.value + B.value
predicted_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(predicted_influence_matrix))
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform'])
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
corrz = []
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
y_test[index]['influence_matrix']))
# Optimization model prediction:
A1 = element['first_influence_matrices']
p = element['individual_performance_hardness_weighted']
A2 = np.row_stack([p, p, p, p])
predicted_influence_matrix = A1 * W1.value + A2 * W2.value + B.value
predicted_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(predicted_influence_matrix))
cr = compute_matrix_err(
influence_matrix, predicted_influence_matrix)
corrz.append(cr)
corrz = np.array(corrz)
plt.hist(corrz[corrz != 0])
###Output
_____no_output_____
###Markdown
With first influence matrix and individual performance (with frob norm)
###Code
sns.heatmap(W1.value);
sns.heatmap(W2.value);
sns.heatmap(B.value);
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform'])
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6273743596380378 +- 0.022857344339732598
uniform: 0.3325318587873968 +- 0.016509363259459668
model: 0.30392078030000086 +- 0.014614658704437995
###Markdown
With dataset itself (not replicating)
###Code
with Timer():
lambdaa = 0.1
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
# X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
# X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
W2 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = utils.make_matrix_row_stochastic(
y_train[index]['influence_matrix'])
A1 = element['first_influence_matrices']
p = element['individual_performance_hardness_weighted']
A2 = np.row_stack([p, p, p, p])
pred_influence_matrix = A1 * W1 + A2 * W2 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
constraints += [pred_influence_matrix >= 0]
constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = utils.make_matrix_row_stochastic(
y_test[index]['influence_matrix'])
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
A1 = element['first_influence_matrices']
p = element['individual_performance_hardness_weighted']
A2 = np.row_stack([p, p, p, p])
predicted_influence_matrix = A1 * W1.value + A2 * W2.value + B.value
predicted_influence_matrix = utils.make_matrix_row_stochastic(predicted_influence_matrix)
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
sns.heatmap(W1.value);
sns.heatmap(W2.value);
sns.heatmap(B.value);
# With dataset itself:
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform'])
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.627369947142661 +- 0.016018037218512353
uniform: 0.33690498301216276 +- 0.020407942132259903
model: 0.30886834958388787 +- 0.012053484282264085
###Markdown
With previous influence matrices and all networks
###Code
with Timer():
runs = 5
lambdaas = [0, 0.01, 0.05, 0.1, 0.2, 0.3, 0.5, 0.9, 1, 2, 5, 10, 100, 1000, 10000]
model_errs = defaultdict(list)
for lambdaa in lambdaas:
print('Lambda: ', lambdaa, '...')
for run in range(runs):
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
W2 = cp.Variable(4, 4)
W3 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
# A1 = element['average_of_previous_influence_matrices']
A1 = element['first_influence_matrices']
A2 = element['reply_duration']
A3 = element['sentiment']
# A4 = element['emotion_arousal']
# A5 = element['emotion_dominance']
# A6 = element['emotion_valence']
pred_influence_matrix = A1 * W1 + A2 * W2 + A3 * W3 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(W2) + cp.norm1(W3) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Optimization model prediction:
A1 = element['first_influence_matrices']
A2 = element['reply_duration']
A3 = element['sentiment']
predicted_influence_matrix = A1 * W1.value + A2 * W2.value + A3 * W3.value + B.value
# predicted_influence_matrix = utils.make_matrix_row_stochastic(predicted_influence_matrix) # << UNCOMMENT IT >>
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
model_errs[lambdaa].append(model_err)
errz = []
for lambdaa in lambdaas:
print(lambdaa, ': ', np.mean(model_errs[lambdaa]), '+-', np.std(model_errs[lambdaa]))
errz.append(np.mean(model_errs[lambdaa]))
###Output
0 : 0.3011743802652413 +- 0.020063083113073005
0.01 : 0.2963133866176929 +- 0.015973161966079022
0.05 : 0.3062668795795276 +- 0.013874774618744515
0.1 : 0.3095999597617864 +- 0.017819090728314457
0.2 : 0.2941185878549716 +- 0.02006738367903435
0.3 : 0.30257457579288827 +- 0.013306889155314812
0.5 : 0.3078848763423549 +- 0.005515771787705399
0.9 : 0.29270645819532487 +- 0.01848981548039053
1 : 0.30146502732172176 +- 0.015082943084012262
2 : 0.30887201679195064 +- 0.020579339692010812
5 : 0.30389845778067637 +- 0.014803297392890165
10 : 0.3096589878225967 +- 0.01617236135673211
100 : 0.3108994226093108 +- 0.015821189229998523
1000 : 0.4881974650919253 +- 0.01133999614957398
10000 : 0.9999999956332914 +- 3.214929424423093e-10
###Markdown
Runs with tunned lambda
###Code
lambdaa = 0.9
runs = 30
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
W2 = cp.Variable(4, 4)
W3 = cp.Variable(4, 4)
# W4 = cp.Variable(4, 4)
# W5 = cp.Variable(4, 4)
# W6 = cp.Variable(4, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
# A1 = element['average_of_previous_influence_matrices']
A1 = element['first_influence_matrices']
A2 = element['reply_duration']
A3 = element['sentiment']
# A4 = element['emotion_arousal']
# A5 = element['emotion_dominance']
# A6 = element['emotion_valence']
pred_influence_matrix = A1 * W1 + A2 * W2 + A3 * W3 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(W2) + cp.norm1(W3) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
# A1 = element['average_of_previous_influence_matrices']
A1 = element['first_influence_matrices']
A2 = element['reply_duration']
A3 = element['sentiment']
# A4 = element['emotion_arousal']
# A5 = element['emotion_dominance']
# A6 = element['emotion_valence']
predicted_influence_matrix = A1 * W1.value + A2 * W2.value + A3 * W3.value + B.value
# predicted_influence_matrix = utils.make_matrix_row_stochastic(predicted_influence_matrix) # << UNCOMMENT IT >>
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
# err += frob_norm_of_difference
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
sum(np.array(uniform_errs) > np.array(model_errs)) / len(model_errs)
plt.plot(np.array(uniform_errs) - np.array(model_errs), '*');
sns.heatmap(W1.value);
sns.heatmap(W2.value);
sns.heatmap(W3.value);
sns.heatmap(B.value);
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
# With the data itself:
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
# plt.hist(model_errs)
# # plt.hist(random_errs)
# plt.hist(uniform_errs)
# # plt.legend(['model', 'random', 'uniform']);
# plt.legend(['model', 'uniform'])
# print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
# print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
# print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6256036346292906 +- 0.02295350141450487
uniform: 0.33512193590092 +- 0.019971630246377763
model: 0.2452193368273235 +- 0.014147432504710407
###Markdown
With text embeddings and the first influence matrix:
###Code
# with Timer():
# runs = 5
# lambdaas = [0, 0.01, 0.05, 0.1, 0.2, 0.3, 0.5, 0.9, 1, 2, 5, 10, 100, 1000, 10000]
# model_errs = defaultdict(list)
# for lambdaa in lambdaas:
# print('Lambda: ', lambdaa, '...')
# for run in range(runs):
# X_train, X_test, y_train, y_test = train_test_split(
# data['X'], data['y'], test_size=test_fraction)
# X_train, y_train = replicate_train_dataset(X_train, y_train)
# # Solving the optimization problem.
# with Timer():
# W1 = cp.Variable(4, 4)
# W2 = cp.Variable(768, 4)
# B = cp.Variable(4, 4)
# constraints = []
# losses = 0
# for index in range(len(X_train)):
# element = X_train[index]
# influence_matrix = y_train[index]['influence_matrix']
# A1 = element['first_influence_matrices']
# A2 = element['content_embedding_matrix']
# pred_influence_matrix = A1 * W1 + A2 * W2 + B
# loss = pred_influence_matrix - influence_matrix
# losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
# regluarization = cp.norm1(W1) + cp.norm1(W2) + cp.norm1(B)
# objective = cp.Minimize(losses + lambdaa * regluarization)
# prob = cp.Problem(objective, constraints)
# result = prob.solve(solver=cp.MOSEK)
# print('It was {} and result was {}'.format(prob.status, result))
# model_err = 0
# for index in range(len(X_test)):
# element = X_test[index]
# influence_matrix = y_test[index]['influence_matrix']
# # Optimization model prediction:
# A1 = element['first_influence_matrices']
# A2 = element['content_embedding_matrix']
# predicted_influence_matrix = A1 * W1.value + A2 * W2.value + B.value
# model_err += compute_matrix_err(
# influence_matrix, predicted_influence_matrix)
# model_err /= len(X_test)
# model_errs[lambdaa].append(model_err)
# errz = []
# for lambdaa in lambdaas:
# print(lambdaa, ': ', np.mean(model_errs[lambdaa]), '+-', np.std(model_errs[lambdaa]))
# errz.append(np.mean(model_errs[lambdaa]))
###Output
_____no_output_____
###Markdown
Runs
###Code
lambdaa = 0.01
runs = 30
model_errs = []
random_errs = []
uniform_errs = []
for run in range(runs):
print('Run', run, '...')
X_train, X_test, y_train, y_test = train_test_split(
data['X'], data['y'], test_size=test_fraction)
X_train, y_train = utils.replicate_networks_in_train_dataset_with_reordering(
X_train, y_train)
# Solving the optimization problem.
with Timer():
W1 = cp.Variable(4, 4)
W2 = cp.Variable(768, 4)
B = cp.Variable(4, 4)
# constraints = []
losses = 0
for index in range(len(X_train)):
element = X_train[index]
influence_matrix = y_train[index]['influence_matrix']
A1 = element['first_influence_matrices']
A2 = element['content_embedding_matrix']
pred_influence_matrix = A1 * W1 + A2 * W2 + B
loss = pred_influence_matrix - influence_matrix
losses += cp.sum_squares(loss)
# constraints += [pred_influence_matrix >= 0]
# constraints += [cp.sum_entries(pred_influence_matrix, axis=1) == 1]
regluarization = cp.norm1(W1) + cp.norm1(W2) + cp.norm1(B)
objective = cp.Minimize(losses + lambdaa * regluarization)
prob = cp.Problem(objective) #, constraints)
result = prob.solve(solver=cp.MOSEK)
print('It was {} and result was {}'.format(prob.status, result))
model_err = 0
random_err = 0
uniform_err = 0
for index in range(len(X_test)):
element = X_test[index]
influence_matrix = y_test[index]['influence_matrix']
# Random model prediction:
pred_random_influence_matrix = np.matrix(utils.make_matrix_row_stochastic(
np.random.rand(4, 4)))
random_err += compute_matrix_err(
influence_matrix, pred_random_influence_matrix)
# Uniform prediction:
pred_uniform_influence_matrix = np.matrix(np.ones((4, 4)) * 0.25)
uniform_err += compute_matrix_err(
influence_matrix, pred_uniform_influence_matrix)
# Optimization model prediction:
A1 = element['first_influence_matrices']
A2 = element['content_embedding_matrix']
predicted_influence_matrix = A1 * W1.value + A2 * W2.value + B.value
model_err += compute_matrix_err(
influence_matrix, predicted_influence_matrix)
model_err /= len(X_test)
random_err /= len(X_test)
uniform_err /= len(X_test)
model_errs.append(model_err)
random_errs.append(random_err)
uniform_errs.append(uniform_err)
plt.hist(model_errs)
plt.hist(random_errs)
plt.hist(uniform_errs)
plt.legend(['model', 'random', 'uniform']);
# plt.legend(['model', 'uniform'])
print('random: {} +- {}'.format(np.mean(random_errs), np.std(random_errs)))
print('uniform: {} +- {}'.format(np.mean(uniform_errs), np.std(uniform_errs)))
print('model: {} +- {}'.format(np.mean(model_errs), np.std(model_errs)));
###Output
random: 0.6269796734190237 +- 0.01995421860943538
uniform: 0.33240827017492997 +- 0.02336559677869273
model: 0.3027936262918508 +- 0.015856241631805136
|
RealDataPractice.ipynb | ###Markdown
Working with a real world data-set using SQL and Python IntroductionThis notebook shows how to work with a real world dataset using SQL and Python. In this lab you will:1. Understand the dataset for Chicago Public School level performance 1. Store the dataset in an Db2 database on IBM Cloud instance1. Retrieve metadata about tables and columns and query data from mixed case columns1. Solve example problems to practice your SQL skills including using built-in database functions Chicago Public Schools - Progress Report Cards (2011-2012) The city of Chicago released a dataset showing all school level performance data used to create School Report Cards for the 2011-2012 school year. The dataset is available from the Chicago Data Portal: https://data.cityofchicago.org/Education/Chicago-Public-Schools-Progress-Report-Cards-2011-/9xs2-f89tThis dataset includes a large number of metrics. Start by familiarizing yourself with the types of metrics in the database: https://data.cityofchicago.org/api/assets/AAD41A13-BE8A-4E67-B1F5-86E711E09D5F?download=true__NOTE__: Do not download the dataset directly from City of Chicago portal. Instead download a more database friendly version from the link below.Now download a static copy of this database and review some of its contents:https://ibm.box.com/shared/static/f9gjvj1gjmxxzycdhplzt01qtz0s7ew7.csv Store the dataset in a TableIn many cases the dataset to be analyzed is available as a .CSV (comma separated values) file, perhaps on the internet. To analyze the data using SQL, it first needs to be stored in the database.While it is easier to read the dataset into a Pandas dataframe and then PERSIST it into the database as we saw in the previous lab, it results in mapping to default datatypes which may not be optimal for SQL querying. For example a long textual field may map to a CLOB instead of a VARCHAR. Now open the Db2 console, open the LOAD tool, Select / Drag the .CSV file for the CHICAGO PUBLIC SCHOOLS dataset and load the dataset into a new table called __SCHOOLS__. Connect to the databaseLet us now load the ipython-sql extension and establish a connection with the database
###Code
%load_ext sql
# Enter the connection string for your Db2 on Cloud database instance below
# %sql ibm_db_sa://my-username:my-password@my-hostname:my-port/my-db-name
%sql ibm_db_sa://username:pwd@host:50000/BLUDB
###Output
(ibm_db_dbi.ProgrammingError) ibm_db_dbi::ProgrammingError: [IBM][CLI Driver] SQL0438N Application raised error or warning with diagnostic text: "Exceeded maximum limit of 5 connections. Connection refused". SQLSTATE=42502\r SQLCODE=-438 (Background on this error at: http://sqlalche.me/e/f405)
Connection info needed in SQLAlchemy format, example:
postgresql://username:password@hostname/dbname
or an existing connection: dict_keys([])
###Markdown
Query the database system catalog to retrieve table metadata You can verify that the table creation was successful by retrieving the list of all tables in your schema and checking whether the SCHOOLS table was created
###Code
# type in your query to retrieve list of all tables in the database for your db2 schema (username)
%sql select TABSCHEMA, TABNAME, CREATE_TIME from SYSCAT.TABLES \
where TABSCHEMA not in ('SYSIBM', 'SYSCAT', 'SYSSTAT', 'SYSIBMADM', 'SYSTOOLS', 'SYSPUBLIC')
###Output
Environment variable $DATABASE_URL not set, and no connect string given.
Connection info needed in SQLAlchemy format, example:
postgresql://username:password@hostname/dbname
or an existing connection: dict_keys([])
###Markdown
Double-click __here__ for a hint<!--In Db2 the system catalog table called SYSCAT.TABLES contains the table metadata--> Double-click __here__ for the solution.<!-- Solution:%sql select TABSCHEMA, TABNAME, CREATE_TIME from SYSCAT.TABLES where TABSCHEMA='YOUR-DB2-USERNAME'or, you can retrieve list of all tables where the schema name is not one of the system created ones:%sql select TABSCHEMA, TABNAME, CREATE_TIME from SYSCAT.TABLES \ where TABSCHEMA not in ('SYSIBM', 'SYSCAT', 'SYSSTAT', 'SYSIBMADM', 'SYSTOOLS', 'SYSPUBLIC') or, just query for a specifc table that you want to verify exists in the database%sql select * from SYSCAT.TABLES where TABNAME = 'SCHOOLS'--> Query the database system catalog to retrieve column metadata The SCHOOLS table contains a large number of columns. How many columns does this table have?
###Code
# type in your query to retrieve the number of columns in the SCHOOLS table
%sql SELECT COUNT(*) FROM syscat.columns WHERE tabname = 'SCHOOLS'
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for a hint<!--In Db2 the system catalog table called SYSCAT.COLUMNS contains the column metadata--> Double-click __here__ for the solution.<!-- Solution:%sql select count(*) from SYSCAT.COLUMNS where TABNAME = 'SCHOOLS'--> Now retrieve the the list of columns in SCHOOLS table and their column type (datatype) and length.
###Code
# type in your query to retrieve all column names in the SCHOOLS table along with their datatypes and length
%sql select COLNAME, TYPENAME, LENGTH from SYSCAT.COLUMNS where TABNAME = 'SCHOOLS'
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for the solution.<!-- Solution:%sql select COLNAME, TYPENAME, LENGTH from SYSCAT.COLUMNS where TABNAME = 'SCHOOLS'or%sql select distinct(NAME), COLTYPE, LENGTH from SYSIBM.SYSCOLUMNS where TBNAME = 'SCHOOLS'--> Questions1. Is the column name for the "SCHOOL ID" attribute in upper or mixed case?1. What is the name of "Community Area Name" column in your table? Does it have spaces?1. Are there any columns in whose names the spaces and paranthesis (round brackets) have been replaced by the underscore character "_"? Problems Problem 1 How many Elementary Schools are in the dataset?
###Code
%sql select count(*) from SCHOOLS where "Elementary, Middle, or High School" = 'ES'
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for a hint<!--Which column specifies the school type e.g. 'ES', 'MS', 'HS'?--> Double-click __here__ for another hint<!--Does the column name have mixed case, spaces or other special characters?If so, ensure you use double quotes around the "Name of the Column"--> Double-click __here__ for the solution.<!-- Solution:%sql select count(*) from SCHOOLS where "Elementary, Middle, or High School" = 'ES'Correct answer: 462--> Problem 2 What is the highest Safety Score?
###Code
%sql select max(SAFETY_SCORE) from SCHOOLS
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for a hint<!--Use the MAX() function--> Double-click __here__ for the solution.<!-- Hint:%sql select MAX(Safety_Score) AS MAX_SAFETY_SCORE from SCHOOLSCorrect answer: 99--> Problem 3 Which schools have highest Safety Score?
###Code
%sql select NAME_OF_SCHOOL, SAFETY_SCORE from SCHOOLS\
where SAFETY_SCORE = (select max(SAFETY_SCORE) from SCHOOLS)
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for the solution.<!-- Solution:In the previous problem we found out that the highest Safety Score is 99, so we can use that as an input in the where clause:%sql select Name_of_School, Safety_Score from SCHOOLS where Safety_Score = 99or, a better way:%sql select Name_of_School, Safety_Score from SCHOOLS where \ Safety_Score= (select MAX(Safety_Score) from SCHOOLS)Correct answer: several schools with with Safety Score of 99.--> Problem 4 What are the top 10 schools with the highest "Average Student Attendance"?
###Code
%sql select NAME_OF_SCHOOL, AVERAGE_STUDENT_ATTENDANCE from SCHOOLS order by AVERAGE_STUDENT_ATTENDANCE desc nulls last limit 10;
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for the solution.<!-- Solution:%sql select Name_of_School, Average_Student_Attendance from SCHOOLS \ order by Average_Student_Attendance desc nulls last limit 10 --> Problem 5 Retrieve the list of 5 Schools with the lowest Average Student Attendance sorted in ascending order based on attendance
###Code
%sql select NAME_OF_SCHOOL, AVERAGE_STUDENT_ATTENDANCE from SCHOOLS\
order by AVERAGE_STUDENT_ATTENDANCE limit 5;
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for the solution.<!-- Solution:%sql SELECT Name_of_School, Average_Student_Attendance \ from SCHOOLS \ order by Average_Student_Attendance \ fetch first 5 rows only--> Problem 6 Now remove the '%' sign from the above result set for Average Student Attendance column
###Code
%sql select NAME_OF_SCHOOL, replace(AVERAGE_STUDENT_ATTENDANCE, '%', '')\
from SCHOOLS order by AVERAGE_STUDENT_ATTENDANCE fetch first 5 rows only
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for a hint<!--Use the REPLACE() function to replace '%' with ''See documentation for this function at:https://www.ibm.com/support/knowledgecenter/en/SSEPGG_10.5.0/com.ibm.db2.luw.sql.ref.doc/doc/r0000843.html--> Double-click __here__ for the solution.<!-- Hint:%sql SELECT Name_of_School, REPLACE(Average_Student_Attendance, '%', '') \ from SCHOOLS \ order by Average_Student_Attendance \ fetch first 5 rows only--> Problem 7 Which Schools have Average Student Attendance lower than 70%?
###Code
%sql select NAME_OF_SCHOOL, AVERAGE_STUDENT_ATTENDANCE from SCHOOLS\
where decimal(replace(AVERAGE_STUDENT_ATTENDANCE, '%','')) <70
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for a hint<!--The datatype of the "Average_Student_Attendance" column is varchar.So you cannot use it as is in the where clause for a numeric comparison.First use the CAST() function to cast it as a DECIMAL or DOUBLEe.g. CAST("Column_Name" as DOUBLE)or simply: DECIMAL("Column_Name")--> Double-click __here__ for another hint<!--Don't forget the '%' age sign needs to be removed before casting--> Double-click __here__ for the solution.<!-- Solution:%sql SELECT Name_of_School, Average_Student_Attendance \ from SCHOOLS \ where CAST ( REPLACE(Average_Student_Attendance, '%', '') AS DOUBLE ) < 70 \ order by Average_Student_Attendance or,%sql SELECT Name_of_School, Average_Student_Attendance \ from SCHOOLS \ where DECIMAL ( REPLACE(Average_Student_Attendance, '%', '') ) < 70 \ order by Average_Student_Attendance--> Problem 8 Get the total College Enrollment for each Community Area
###Code
%sql select Community_Area_Name, sum(College_Enrollment) AS TOTAL_ENROLLMENT \
from SCHOOLS \
group by Community_Area_Name
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
###Markdown
Double-click __here__ for a hint<!--Verify the exact name of the Enrollment column in the databaseUse the SUM() function to add up the Enrollments for each Community Area--> Double-click __here__ for another hint<!--Don't forget to group by the Community Area--> Double-click __here__ for the solution.<!-- Solution:%sql select Community_Area_Name, sum(College_Enrollment) AS TOTAL_ENROLLMENT \ from SCHOOLS \ group by Community_Area_Name --> Problem 9 Get the 5 Community Areas with the least total College Enrollment sorted in ascending order
###Code
%sql select Community_Area_Name, sum(College_Enrollment) AS TOTAL_ENROLLMENT \
from SCHOOLS \
group by Community_Area_Name \
order by TOTAL_ENROLLMENT asc \
fetch first 5 rows only
###Output
* ibm_db_sa://clk37870:***@dashdb-txn-sbox-yp-lon02-01.services.eu-gb.bluemix.net:50000/BLUDB
Done.
|
exercise/chap_6/7_dt_moons.ipynb | ###Markdown
Important Points* Created using make_moons function in sklean a synthetic dataset* Applied GridSearchCV with params max_leaf_nodes and max_depth| Model | Training | Validation ||--------------|--------------|-------------|| Decison Tree | 86.28% | 85.68% |
###Code
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
np.random.seed(42)
%matplotlib inline
# Creating data and train-test split
from sklearn.datasets import make_moons
from sklearn.model_selection import train_test_split
x, y = make_moons(n_samples=10000, noise=0.4, random_state=True)
x_train, x_valid, y_train, y_valid = train_test_split(x, y, random_state=42, test_size=0.25)
print(x_train.shape, y_train.shape)
print(x_valid.shape, y_valid.shape)
# No of classes
np.unique(y_train)
from sklearn.metrics import accuracy_score
def score(y, y_pred, train=False):
if train:
print("Training accuracy: ", accuracy_score(y, y_pred))
else:
print("Validation accuracy: ", accuracy_score(y, y_pred))
from sklearn.tree import DecisionTreeClassifier
dt = DecisionTreeClassifier()
dt.fit(x_train, y_train)
# The model overfits since no regularization had been applied
score(y_valid, dt.predict(x_valid))
score(y_train, dt.predict(x_train), True)
from sklearn.model_selection import GridSearchCV
params = {"max_leaf_nodes": list(range(2, 50)), "max_depth": list(range(8, 12))}
grid_search = GridSearchCV(DecisionTreeClassifier(random_state=42), verbose=1, param_grid=params, cv=3)
%time
grid_search.fit(x_train, y_train)
grid_search.best_estimator_
# Since refit=True, model already trained on complete x_train
score(y_train, grid_search.predict(x_train), True)
# Validation score
score(y_valid, grid_search.predict(x_valid))
###Output
Training accuracy: 0.8628
Validation accuracy: 0.8568
|
src/databricks/notebooks/traffic-camera-speeddetection.py.ipynb | ###Markdown
Load variables from key vault
###Code
kv_scope = 'key-vault-secret'
# Variables
storage_account_name = dbutils.secrets.get(scope=kv_scope, key='traffic-storage-accountname')
storage_account_access_key = dbutils.secrets.get(scope=kv_scope, key='traffic-storage-accountkey')
eventgrid_accesskey = dbutils.secrets.get(scope=kv_scope, key='traffic-eventgrid-accesskey')
eventgrid_topic = dbutils.secrets.get(scope=kv_scope, key='traffic-eventgrid-topicendpoint')
###Output
_____no_output_____
###Markdown
Mounting the segment configuration json from blob- Using the mount functionality to load the blob file
###Code
mount_name = 'traffic-config'
to_be_mounted = True
mounts = dbutils.fs.ls('/mnt/')
for mnt in mounts:
if mnt.name.startswith(mount_name):
to_be_mounted = False
if to_be_mounted:
dbutils.fs.mount(
source = 'wasbs://traffic-config@' + storage_account_name + '.blob.core.windows.net',
mount_point = '/mnt/' + mount_name,
extra_configs = {'fs.azure.account.key.' + storage_account_name + '.blob.core.windows.net':storage_account_access_key})
else:
print('Traffic config already mounted')
###Output
_____no_output_____
###Markdown
Parsing segment configuration- Reading the json file (`multiLine=True` !!)- Adding calculated field for maximum duration (`(distance / speedlimit) * 3.6`), where 3.6 is coming from meters/second- Only returning the relevant fields for the calculation query
###Code
segment_config = spark.read.json('/mnt/' + mount_name, multiLine=True) \
.withColumn('TrajectId', col('segmentId')) \
.withColumn('MinDuration', ((col('cameraDistance') / col('speedLimit')) * 3.6)) \
.select('TrajectId', 'MinDuration', 'CameraDistance', 'SpeedLimit') \
display(segment_config)
timestamp_from = datetime.utcnow() - timedelta(hours=0, minutes=20)
delta_src_table_name = 'CameraTelemetry' + datetime.today().strftime('%Y%m%d')
delta_dest_table_name = 'SpeedMeasurements' + datetime.today().strftime('%Y%m%d')
cameraStream = spark.readStream.format('delta') \
.table(delta_src_table_name) \
.withWatermark('EventTime', '10 seconds')
###Output
_____no_output_____
###Markdown
Query that measures time difference per licenseplate- Loading data from the delta table- Grouping by license plate and traject- Taking count, earliest timestamp and latest timestamp- Adding calculated field (max-min) for duration- Selecting relevant fields as output
###Code
duration_calculation = cameraStream \
.groupBy('TrajectId', 'LicensePlate', 'Make', 'Country') \
.agg(count('*').alias('count'), min('EventTime').alias('firstevent'), max('EventTime').alias('lastevent')) \
.withColumn('duration', col('lastevent').cast(LongType())-col('firstevent').cast(LongType())) \
.where((col('duration') > 0) & (col('count') >= 2)) \
.select('TrajectId','LicensePlate', 'Make', 'Country', 'FirstEvent', 'LastEvent', 'Count', 'Duration')
###Output
_____no_output_____
###Markdown
Join results with traject information and detect speeding cars- Join on TrajectId- Select cars with duration that is below the minimum duration of the traject- Add a calculated column for speed
###Code
speed_measurements_df = duration_calculation.join(segment_config, 'TrajectId') \
.withColumn('speed', ((col('CameraDistance') / col('duration')) * 3.6)) \
.select('TrajectId', 'LicensePlate', 'Speed', 'Make', 'Country', 'LastEvent', 'SpeedLimit', 'Duration') \
.withWatermark('LastEvent', '5 seconds') \
.writeStream \
.format('delta') \
.outputMode('complete') \
.option('checkpointLocation', '/data/' + delta_dest_table_name + 'cp/_checkpoints/data_file') \
.table(delta_dest_table_name)
speed_tickets_df = duration_calculation.join(segment_config, 'TrajectId') \
.withColumn('Subject', concat(col('TrajectId'), lit('/'), col('LicensePlate'))) \
.withColumn('speed', ((col('CameraDistance') / col('duration')) * 3.6)) \
.where(col('duration') < col('MinDuration')) \
.select('TrajectId', 'LicensePlate', 'Speed', 'Duration', 'Subject', 'SpeedLimit') \
.writeStream.foreach(eg.EventGridSinkWriter(eventgrid_topic, eventgrid_accesskey, 'SpeedingCarDetected')) \
.outputMode('update') \
.start()
#dbutils.fs.unmount('/mnt/' + mount_name)
###Output
_____no_output_____ |
CS224W_Colab1.ipynb | ###Markdown
**CS224W - Colab 1** In this Colab, we will write a full pipeline for **learning node embeddings**.We will go through the following 3 steps.To start, we will load a classic graph in network science, the [Karate Club Network](https://en.wikipedia.org/wiki/Zachary%27s_karate_club). We will explore multiple graph statistics for that graph.We will then work together to transform the graph structure into a PyTorch tensor, so that we can perform machine learning over the graph.Finally, we will finish the first learning algorithm on graphs: a node embedding model. For simplicity, our model here is simpler than DeepWalk / node2vec algorithms taught in the lecture. But it's still rewarding and challenging, as we will write it from scratch via PyTorch.Now let's get started!**Note**: Make sure to **sequentially run all the cells**, so that the intermediate variables / packages will carry over to the next cell 1 Graph BasicsTo start, we will load a classic graph in network science, the [Karate Club Network](https://en.wikipedia.org/wiki/Zachary%27s_karate_club). We will explore multiple graph statistics for that graph. SetupWe will heavily use NetworkX in this Colab.
###Code
import networkx as nx
###Output
_____no_output_____
###Markdown
Zachary's karate club networkThe [Karate Club Network](https://en.wikipedia.org/wiki/Zachary%27s_karate_club) is a graph describes a social network of 34 members of a karate club and documents links between members who interacted outside the club.
###Code
G = nx.karate_club_graph()
# G is an undirected graph
type(G)
# Visualize the graph
nx.draw(G, with_labels = True)
###Output
_____no_output_____
###Markdown
Question 1: What is the average degree of the karate club network? (5 Points)
###Code
def average_degree(num_edges, num_nodes):
# TODO: Implement this function that takes number of edges
# and number of nodes, and returns the average node degree of
# the graph. Round the result to nearest integer (for example
# 3.3 will be rounded to 3 and 3.7 will be rounded to 4)
avg_degree = 0
############# Your code here ############
avg_degree = round(2 * num_edges / num_nodes)
#########################################
return avg_degree
num_edges = G.number_of_edges()
num_nodes = G.number_of_nodes()
avg_degree = average_degree(num_edges, num_nodes)
print("Average degree of karate club network is {}".format(avg_degree))
###Output
Average degree of karate club network is 5
###Markdown
Question 2: What is the average clustering coefficient of the karate club network? (5 Points)
###Code
def average_clustering_coefficient(G):
# TODO: Implement this function that takes a nx.Graph
# and returns the average clustering coefficient. Round
# the result to 2 decimal places (for example 3.333 will
# be rounded to 3.33 and 3.7571 will be rounded to 3.76)
avg_cluster_coef = 0
############# Your code here ############
## Note:
## 1: Please use the appropriate NetworkX clustering function
avg_cluster_coef = round(nx.average_clustering(G), 2)
#########################################
return avg_cluster_coef
avg_cluster_coef = average_clustering_coefficient(G)
print("Average clustering coefficient of karate club network is {}".format(avg_cluster_coef))
###Output
Average clustering coefficient of karate club network is 0.57
###Markdown
Question 3: What is the PageRank value for node 0 (node with id 0) after one PageRank iteration? (5 Points)Please complete the code block by implementing the PageRank equation: $r_j = \sum_{i \rightarrow j} \beta \frac{r_i}{d_i} + (1 - \beta) \frac{1}{N}$
###Code
def one_iter_pagerank(G, beta, r0, node_id):
# TODO: Implement this function that takes a nx.Graph, beta, r0 and node id.
# The return value r1 is one interation PageRank value for the input node.
# Please round r1 to 2 decimal places.
r1 = 0
############# Your code here ############
## Note:
## 1: You should not use nx.pagerank
for neighbor in G.neighbors(node_id):
r1 += beta * r0 / G.degree[neighbor]
r1 += (1 - beta) / G.number_of_nodes()
r1 = round(r1, 2)
#########################################
return r1
beta = 0.8
r0 = 1 / G.number_of_nodes()
node = 0
r1 = one_iter_pagerank(G, beta, r0, node)
print("The PageRank value for node 0 after one iteration is {}".format(r1))
###Output
The PageRank value for node 0 after one iteration is 0.13
###Markdown
Question 4: What is the (raw) closeness centrality for the karate club network node 5? (5 Points)The equation for closeness centrality is $c(v) = \frac{1}{\sum_{u \neq v}\text{shortest path length between } u \text{ and } v}$
###Code
def closeness_centrality(G, node=5):
# TODO: Implement the function that calculates closeness centrality
# for a node in karate club network. G is the input karate club
# network and node is the node id in the graph. Please round the
# closeness centrality result to 2 decimal places.
closeness = 0
## Note:
## 1: You can use networkx closeness centrality function.
## 2: Notice that networkx closeness centrality returns the normalized
## closeness directly, which is different from the raw (unnormalized)
## one that we learned in the lecture.
normalized_closeness = nx.closeness_centrality(G, u=node)
closeness = normalized_closeness / (len(nx.node_connected_component(G, node)) - 1)
closeness = round(closeness, 2)
#########################################
return closeness
node = 5
closeness = closeness_centrality(G, node=node)
print("The node 5 has closeness centrality {}".format(closeness))
###Output
The node 5 has closeness centrality 0.01
###Markdown
2 Graph to TensorWe will then work together to transform the graph $G$ into a PyTorch tensor, so that we can perform machine learning over the graph. SetupCheck if PyTorch is properly installed
###Code
import torch
print(torch.__version__)
###Output
1.10.0+cu111
###Markdown
PyTorch tensor basicsWe can generate PyTorch tensor with all zeros, ones or random values.
###Code
# Generate 3 x 4 tensor with all ones
ones = torch.ones(3, 4)
print(ones)
# Generate 3 x 4 tensor with all zeros
zeros = torch.zeros(3, 4)
print(zeros)
# Generate 3 x 4 tensor with random values on the interval [0, 1)
random_tensor = torch.rand(3, 4)
print(random_tensor)
# Get the shape of the tensor
print(ones.shape)
###Output
tensor([[1., 1., 1., 1.],
[1., 1., 1., 1.],
[1., 1., 1., 1.]])
tensor([[0., 0., 0., 0.],
[0., 0., 0., 0.],
[0., 0., 0., 0.]])
tensor([[0.6379, 0.4626, 0.4569, 0.9950],
[0.4649, 0.5162, 0.2455, 0.5794],
[0.2019, 0.4957, 0.6122, 0.1658]])
torch.Size([3, 4])
###Markdown
PyTorch tensor contains elements for a single data type, the `dtype`.
###Code
# Create a 3 x 4 tensor with all 32-bit floating point zeros
zeros = torch.zeros(3, 4, dtype=torch.float32)
print(zeros.dtype)
# Change the tensor dtype to 64-bit integer
zeros = zeros.type(torch.long)
print(zeros.dtype)
###Output
torch.float32
torch.int64
###Markdown
Question 5: Get the edge list of the karate club network and transform it into `torch.LongTensor`. What is the `torch.sum` value of `pos_edge_index` tensor? (10 Points)
###Code
def graph_to_edge_list(G):
# TODO: Implement the function that returns the edge list of
# an nx.Graph. The returned edge_list should be a list of tuples
# where each tuple is a tuple representing an edge connected
# by two nodes.
edge_list = []
############# Your code here ############
edge_list = list(list(G.edges()))
#########################################
return edge_list
def edge_list_to_tensor(edge_list):
# TODO: Implement the function that transforms the edge_list to
# tensor. The input edge_list is a list of tuples and the resulting
# tensor should have the shape [2 x len(edge_list)].
edge_index = torch.tensor([])
############# Your code here ############
edge_index = torch.tensor(edge_list).T
#########################################
return edge_index
pos_edge_list = graph_to_edge_list(G)
pos_edge_index = edge_list_to_tensor(pos_edge_list)
print("The pos_edge_index tensor has shape {}".format(pos_edge_index.shape))
print("The pos_edge_index tensor has sum value {}".format(torch.sum(pos_edge_index)))
###Output
The pos_edge_index tensor has shape torch.Size([2, 78])
The pos_edge_index tensor has sum value 2535
###Markdown
Question 6: Please implement following function that samples negative edges. Then answer which edges (edge_1 to edge_5) can be potential negative edges in the karate club network? (10 Points)
###Code
import random
def sample_negative_edges(G, num_neg_samples):
# TODO: Implement the function that returns a list of negative edges.
# The number of sampled negative edges is num_neg_samples. You do not
# need to consider the corner case when the number of possible negative edges
# is less than num_neg_samples. It should be ok as long as your implementation
# works on the karate club network. In this implementation, self loops should
# not be considered as either a positive or negative edge. Also, notice that
# the karate club network is an undirected graph, if (0, 1) is a positive
# edge, do you think (1, 0) can be a negative one?
neg_edge_list = []
############# Your code here ############
pos_set = set(G.edges())
visited_set = set()
node_list = list(G.nodes())
random.shuffle(node_list)
for n_i in node_list:
for n_j in node_list:
if n_i == n_j \
or (n_i,n_j) in pos_set or (n_j,n_i) in pos_set \
or (n_i,n_j) in visited_set or (n_j, n_i) is visited_set:
continue
neg_edge_list.append((n_i,n_j))
visited_set.add((n_i,n_j))
visited_set.add((n_j,n_i))
if len(neg_edge_list) == num_neg_samples:
return neg_edge_list
#########################################
return neg_edge_list
# Sample 78 negative edges
neg_edge_list = sample_negative_edges(G, len(pos_edge_list))
# Transform the negative edge list to tensor
neg_edge_index = edge_list_to_tensor(neg_edge_list)
print("The neg_edge_index tensor has shape {}".format(neg_edge_index.shape))
# Which of following edges can be negative ones?
edge_1 = (7, 1)
edge_2 = (1, 33)
edge_3 = (33, 22)
edge_4 = (0, 4)
edge_5 = (4, 2)
############# Your code here ############
## Note:
## 1: For each of the 5 edges, print whether it can be negative edge
def is_neg_edge(edge):
return not(edge in pos_edge_list or (edge[1], edge[0]) in pos_edge_list)
print(is_neg_edge(edge_1))
print(is_neg_edge(edge_2))
print(is_neg_edge(edge_3))
print(is_neg_edge(edge_4))
print(is_neg_edge(edge_5))
#########################################
###Output
The neg_edge_index tensor has shape torch.Size([2, 78])
False
True
False
False
True
###Markdown
3 Node Emebedding LearningFinally, we will finish the first learning algorithm on graphs: a node embedding model. Setup
###Code
import torch
import torch.nn as nn
import matplotlib.pyplot as plt
from sklearn.decomposition import PCA
print(torch.__version__)
###Output
1.10.0+cu111
###Markdown
To write our own node embedding learning methods, we'll heavily use the [`nn.Embedding`](https://pytorch.org/docs/stable/generated/torch.nn.Embedding.html) module in PyTorch. Let's see how to use `nn.Embedding`:
###Code
# Initialize an embedding layer
# Suppose we want to have embedding for 4 items (e.g., nodes)
# Each item is represented with 8 dimensional vector
emb_sample = nn.Embedding(num_embeddings=4, embedding_dim=8)
print('Sample embedding layer: {}'.format(emb_sample))
###Output
Sample embedding layer: Embedding(4, 8)
###Markdown
We can select items from the embedding matrix, by using Tensor indices
###Code
# Select an embedding in emb_sample
id = torch.LongTensor([1])
print(emb_sample(id))
# Select multiple embeddings
ids = torch.LongTensor([1, 3])
print(emb_sample(ids))
# Get the shape of the embedding weight matrix
shape = emb_sample.weight.data.shape
print(shape)
# Overwrite the weight to tensor with all ones
emb_sample.weight.data = torch.ones(shape)
# Let's check if the emb is indeed initilized
ids = torch.LongTensor([0, 3])
print(emb_sample(ids))
###Output
tensor([[ 0.6998, 0.2015, -0.2159, 0.3704, -0.0126, 2.8423, -0.4166, -0.2715]],
grad_fn=<EmbeddingBackward0>)
tensor([[ 0.6998, 0.2015, -0.2159, 0.3704, -0.0126, 2.8423, -0.4166, -0.2715],
[-1.7620, 0.4619, -1.7589, -0.1419, -1.0391, 0.5771, 0.7422, -0.4567]],
grad_fn=<EmbeddingBackward0>)
torch.Size([4, 8])
tensor([[1., 1., 1., 1., 1., 1., 1., 1.],
[1., 1., 1., 1., 1., 1., 1., 1.]], grad_fn=<EmbeddingBackward0>)
###Markdown
Now, it's your time to create node embedding matrix for the graph we have!- We want to have **16 dimensional** vector for each node in the karate club network.- We want to initalize the matrix under **uniform distribution**, in the range of $[0, 1)$. We suggest you using [`torch.rand`](https://pytorch.org/docs/stable/generated/torch.rand.html).
###Code
# Please do not change / reset the random seed
torch.manual_seed(1)
def create_node_emb(num_node=34, embedding_dim=16):
# TODO: Implement this function that will create the node embedding matrix.
# A torch.nn.Embedding layer will be returned. You do not need to change
# the values of num_node and embedding_dim. The weight matrix of returned
# layer should be initialized under uniform distribution.
emb = None
############# Your code here ############
emb = nn.Embedding(num_nodes, embedding_dim)
emb.weight.data = torch.rand(num_nodes, embedding_dim)
#########################################
return emb
emb = create_node_emb()
ids = torch.LongTensor([0, 3])
# Print the embedding layer
print("Embedding: {}".format(emb))
# An example that gets the embeddings for node 0 and 3
print(emb(ids))
###Output
Embedding: Embedding(34, 16)
tensor([[0.2114, 0.7335, 0.1433, 0.9647, 0.2933, 0.7951, 0.5170, 0.2801, 0.8339,
0.1185, 0.2355, 0.5599, 0.8966, 0.2858, 0.1955, 0.1808],
[0.7486, 0.6546, 0.3843, 0.9820, 0.6012, 0.3710, 0.4929, 0.9915, 0.8358,
0.4629, 0.9902, 0.7196, 0.2338, 0.0450, 0.7906, 0.9689]],
grad_fn=<EmbeddingBackward0>)
###Markdown
Visualize the initial node embeddingsOne good way to understand an embedding matrix, is to visualize it in a 2D space.Here, we have implemented an embedding visualization function for you.We first do PCA to reduce the dimensionality of embeddings to a 2D space.Then we visualize each point, colored by the community it belongs to.
###Code
def visualize_emb(emb):
X = emb.weight.data.numpy()
pca = PCA(n_components=2)
components = pca.fit_transform(X)
plt.figure(figsize=(6, 6))
club1_x = []
club1_y = []
club2_x = []
club2_y = []
for node in G.nodes(data=True):
if node[1]['club'] == 'Mr. Hi':
club1_x.append(components[node[0]][0])
club1_y.append(components[node[0]][1])
else:
club2_x.append(components[node[0]][0])
club2_y.append(components[node[0]][1])
plt.scatter(club1_x, club1_y, color="red", label="Mr. Hi")
plt.scatter(club2_x, club2_y, color="blue", label="Officer")
plt.legend()
plt.show()
# Visualize the initial random embeddding
visualize_emb(emb)
###Output
_____no_output_____
###Markdown
Question 7: Training the embedding! What is the best performance you can get? Please report both the best loss and accuracy on Gradescope. (20 Points)We want to optimize our embeddings for the task of classifying edges as positive or negative. Given an edge and the embeddings for each node, the dot product of the embeddings, followed by a sigmoid, should give us the likelihood of that edge being either positive (output of sigmoid > 0.5) or negative (output of sigmoid < 0.5).Note that we're using the functions you wrote in the previous questions, _as well as the variables initialized in previous cells_. If you're running into issues, make sure your answers to questions 1-6 are correct.
###Code
from torch.optim import SGD
import torch.nn as nn
import numpy as np
def accuracy(pred, label):
# TODO: Implement the accuracy function. This function takes the
# pred tensor (the resulting tensor after sigmoid) and the label
# tensor (torch.LongTensor). Predicted value greater than 0.5 will
# be classified as label 1. Else it will be classified as label 0.
# The returned accuracy should be rounded to 4 decimal places.
# For example, accuracy 0.82956 will be rounded to 0.8296.
accu = 0.0
############# Your code here ############
pred = [1 if item>0.5 else 0 for item in pred]
num_match = (np.array(pred) == np.array(train_label)).sum()
accu = num_match / len(train_label)
accu = round(accu, 4)
#########################################
return accu
def train(emb, loss_fn, sigmoid, train_label, train_edge):
# TODO: Train the embedding layer here. You can also change epochs and
# learning rate. In general, you need to implement:
# (1) Get the embeddings of the nodes in train_edge
# (2) Dot product the embeddings between each node pair
# (3) Feed the dot product result into sigmoid
# (4) Feed the sigmoid output into the loss_fn
# (5) Print both loss and accuracy of each epoch
# (6) Update the embeddings using the loss and optimizer
# (as a sanity check, the loss should decrease during training)
epochs = 500
learning_rate = 0.1
optimizer = SGD(emb.parameters(), lr=learning_rate, momentum=0.9)
for i in range(epochs):
############# Your code here ############
optimizer.zero_grad()
# (1) Get the embeddings of the nodes in train_edge
emb_set_u = emb(train_edge[0])
emb_set_v = emb(train_edge[1])
# (2) Dot product the embeddings between each node pair
dot_prod = torch.sum(emb_set_u * emb_set_v, dim=-1)
# (3) Feed the dot product result into sigmoid
sig = sigmoid(dot_prod)
# (4) Feed the sigmoid output into the loss_fn
loss = loss_fn(sig, train_label)
loss.backward() # Derive gradients.
optimizer.step() # Update parameters based on gradients.
print(f"Loss for Epoch {i}: {loss}")
print(f"Accuracy for is Epoch {i}: {accuracy(sig, train_label)}")
print()
#########################################
loss_fn = nn.BCELoss()
sigmoid = nn.Sigmoid()
print(pos_edge_index.shape)
# Generate the positive and negative labels
pos_label = torch.ones(pos_edge_index.shape[1], )
neg_label = torch.zeros(neg_edge_index.shape[1], )
# Concat positive and negative labels into one tensor
train_label = torch.cat([pos_label, neg_label], dim=0)
# Concat positive and negative edges into one tensor
# Since the network is very small, we do not split the edges into val/test sets
train_edge = torch.cat([pos_edge_index, neg_edge_index], dim=1)
print(train_edge.shape)
train(emb, loss_fn, sigmoid, train_label, train_edge)
###Output
torch.Size([2, 78])
torch.Size([2, 156])
Loss for Epoch 0: 2.25661301612854
Accuracy for is Epoch 0: 0.5
Loss for Epoch 1: 2.2170355319976807
Accuracy for is Epoch 1: 0.5
Loss for Epoch 2: 2.1426429748535156
Accuracy for is Epoch 2: 0.5
Loss for Epoch 3: 2.038400411605835
Accuracy for is Epoch 3: 0.5
Loss for Epoch 4: 1.9093924760818481
Accuracy for is Epoch 4: 0.5
Loss for Epoch 5: 1.7607818841934204
Accuracy for is Epoch 5: 0.5
Loss for Epoch 6: 1.5978357791900635
Accuracy for is Epoch 6: 0.5
Loss for Epoch 7: 1.4259817600250244
Accuracy for is Epoch 7: 0.5
Loss for Epoch 8: 1.2508485317230225
Accuracy for is Epoch 8: 0.5
Loss for Epoch 9: 1.0782299041748047
Accuracy for is Epoch 9: 0.5
Loss for Epoch 10: 0.9139243960380554
Accuracy for is Epoch 10: 0.5
Loss for Epoch 11: 0.763421893119812
Accuracy for is Epoch 11: 0.5192
Loss for Epoch 12: 0.631392776966095
Accuracy for is Epoch 12: 0.5705
Loss for Epoch 13: 0.521044909954071
Accuracy for is Epoch 13: 0.6474
Loss for Epoch 14: 0.4336003065109253
Accuracy for is Epoch 14: 0.6987
Loss for Epoch 15: 0.36819779872894287
Accuracy for is Epoch 15: 0.8077
Loss for Epoch 16: 0.32231464982032776
Accuracy for is Epoch 16: 0.8654
Loss for Epoch 17: 0.2925078570842743
Accuracy for is Epoch 17: 0.9167
Loss for Epoch 18: 0.27514827251434326
Accuracy for is Epoch 18: 0.9103
Loss for Epoch 19: 0.26692166924476624
Accuracy for is Epoch 19: 0.9103
Loss for Epoch 20: 0.2650585174560547
Accuracy for is Epoch 20: 0.9038
Loss for Epoch 21: 0.26737043261528015
Accuracy for is Epoch 21: 0.8974
Loss for Epoch 22: 0.2721913754940033
Accuracy for is Epoch 22: 0.8974
Loss for Epoch 23: 0.27828744053840637
Accuracy for is Epoch 23: 0.8974
Loss for Epoch 24: 0.28476765751838684
Accuracy for is Epoch 24: 0.8974
Loss for Epoch 25: 0.29100707173347473
Accuracy for is Epoch 25: 0.8974
Loss for Epoch 26: 0.2965843379497528
Accuracy for is Epoch 26: 0.8974
Loss for Epoch 27: 0.3012319803237915
Accuracy for is Epoch 27: 0.8974
Loss for Epoch 28: 0.30479806661605835
Accuracy for is Epoch 28: 0.8974
Loss for Epoch 29: 0.3072158992290497
Accuracy for is Epoch 29: 0.8974
Loss for Epoch 30: 0.30848127603530884
Accuracy for is Epoch 30: 0.8974
Loss for Epoch 31: 0.308634877204895
Accuracy for is Epoch 31: 0.8974
Loss for Epoch 32: 0.3077490031719208
Accuracy for is Epoch 32: 0.8974
Loss for Epoch 33: 0.30591732263565063
Accuracy for is Epoch 33: 0.8974
Loss for Epoch 34: 0.3032471537590027
Accuracy for is Epoch 34: 0.8974
Loss for Epoch 35: 0.29985353350639343
Accuracy for is Epoch 35: 0.8974
Loss for Epoch 36: 0.2958546280860901
Accuracy for is Epoch 36: 0.8974
Loss for Epoch 37: 0.2913680970668793
Accuracy for is Epoch 37: 0.8974
Loss for Epoch 38: 0.28650835156440735
Accuracy for is Epoch 38: 0.8974
Loss for Epoch 39: 0.28138408064842224
Accuracy for is Epoch 39: 0.8974
Loss for Epoch 40: 0.2760966718196869
Accuracy for is Epoch 40: 0.8974
Loss for Epoch 41: 0.2707386910915375
Accuracy for is Epoch 41: 0.8974
Loss for Epoch 42: 0.265392929315567
Accuracy for is Epoch 42: 0.8974
Loss for Epoch 43: 0.26013126969337463
Accuracy for is Epoch 43: 0.8974
Loss for Epoch 44: 0.25501468777656555
Accuracy for is Epoch 44: 0.8974
Loss for Epoch 45: 0.2500925362110138
Accuracy for is Epoch 45: 0.8974
Loss for Epoch 46: 0.24540305137634277
Accuracy for is Epoch 46: 0.8974
Loss for Epoch 47: 0.24097339808940887
Accuracy for is Epoch 47: 0.8974
Loss for Epoch 48: 0.23682045936584473
Accuracy for is Epoch 48: 0.9038
Loss for Epoch 49: 0.23295140266418457
Accuracy for is Epoch 49: 0.9103
Loss for Epoch 50: 0.22936496138572693
Accuracy for is Epoch 50: 0.9103
Loss for Epoch 51: 0.226052388548851
Accuracy for is Epoch 51: 0.9103
Loss for Epoch 52: 0.22299876809120178
Accuracy for is Epoch 52: 0.9103
Loss for Epoch 53: 0.22018428146839142
Accuracy for is Epoch 53: 0.9103
Loss for Epoch 54: 0.2175855189561844
Accuracy for is Epoch 54: 0.9231
Loss for Epoch 55: 0.21517665684223175
Accuracy for is Epoch 55: 0.9231
Loss for Epoch 56: 0.212930828332901
Accuracy for is Epoch 56: 0.9231
Loss for Epoch 57: 0.21082088351249695
Accuracy for is Epoch 57: 0.9295
Loss for Epoch 58: 0.20882073044776917
Accuracy for is Epoch 58: 0.9295
Loss for Epoch 59: 0.20690567791461945
Accuracy for is Epoch 59: 0.9359
Loss for Epoch 60: 0.2050534039735794
Accuracy for is Epoch 60: 0.9359
Loss for Epoch 61: 0.2032441347837448
Accuracy for is Epoch 61: 0.9359
Loss for Epoch 62: 0.201461061835289
Accuracy for is Epoch 62: 0.9359
Loss for Epoch 63: 0.19969037175178528
Accuracy for is Epoch 63: 0.9359
Loss for Epoch 64: 0.19792133569717407
Accuracy for is Epoch 64: 0.9359
Loss for Epoch 65: 0.19614599645137787
Accuracy for is Epoch 65: 0.9359
Loss for Epoch 66: 0.19435913860797882
Accuracy for is Epoch 66: 0.9359
Loss for Epoch 67: 0.19255773723125458
Accuracy for is Epoch 67: 0.9359
Loss for Epoch 68: 0.1907408982515335
Accuracy for is Epoch 68: 0.9359
Loss for Epoch 69: 0.18890917301177979
Accuracy for is Epoch 69: 0.9359
Loss for Epoch 70: 0.18706448376178741
Accuracy for is Epoch 70: 0.9359
Loss for Epoch 71: 0.18520960211753845
Accuracy for is Epoch 71: 0.9359
Loss for Epoch 72: 0.18334777653217316
Accuracy for is Epoch 72: 0.9359
Loss for Epoch 73: 0.18148262798786163
Accuracy for is Epoch 73: 0.9359
Loss for Epoch 74: 0.17961779236793518
Accuracy for is Epoch 74: 0.9423
Loss for Epoch 75: 0.1777566820383072
Accuracy for is Epoch 75: 0.9423
Loss for Epoch 76: 0.17590248584747314
Accuracy for is Epoch 76: 0.9423
Loss for Epoch 77: 0.17405793070793152
Accuracy for is Epoch 77: 0.9423
Loss for Epoch 78: 0.17222529649734497
Accuracy for is Epoch 78: 0.9423
Loss for Epoch 79: 0.17040641605854034
Accuracy for is Epoch 79: 0.9423
Loss for Epoch 80: 0.1686026006937027
Accuracy for is Epoch 80: 0.9423
Loss for Epoch 81: 0.16681469976902008
Accuracy for is Epoch 81: 0.9423
Loss for Epoch 82: 0.1650431901216507
Accuracy for is Epoch 82: 0.9423
Loss for Epoch 83: 0.1632881611585617
Accuracy for is Epoch 83: 0.9423
Loss for Epoch 84: 0.16154944896697998
Accuracy for is Epoch 84: 0.9423
Loss for Epoch 85: 0.1598266065120697
Accuracy for is Epoch 85: 0.9423
Loss for Epoch 86: 0.1581190675497055
Accuracy for is Epoch 86: 0.9423
Loss for Epoch 87: 0.15642611682415009
Accuracy for is Epoch 87: 0.9423
Loss for Epoch 88: 0.15474699437618256
Accuracy for is Epoch 88: 0.9423
Loss for Epoch 89: 0.15308091044425964
Accuracy for is Epoch 89: 0.9423
Loss for Epoch 90: 0.15142713487148285
Accuracy for is Epoch 90: 0.9423
Loss for Epoch 91: 0.1497848927974701
Accuracy for is Epoch 91: 0.9487
Loss for Epoch 92: 0.14815357327461243
Accuracy for is Epoch 92: 0.9487
Loss for Epoch 93: 0.14653262495994568
Accuracy for is Epoch 93: 0.9487
Loss for Epoch 94: 0.14492157101631165
Accuracy for is Epoch 94: 0.9487
Loss for Epoch 95: 0.1433200240135193
Accuracy for is Epoch 95: 0.9487
Loss for Epoch 96: 0.1417277455329895
Accuracy for is Epoch 96: 0.9487
Loss for Epoch 97: 0.14014455676078796
Accuracy for is Epoch 97: 0.9679
Loss for Epoch 98: 0.13857032358646393
Accuracy for is Epoch 98: 0.9679
Loss for Epoch 99: 0.13700509071350098
Accuracy for is Epoch 99: 0.9679
Loss for Epoch 100: 0.13544891774654388
Accuracy for is Epoch 100: 0.9679
Loss for Epoch 101: 0.13390189409255981
Accuracy for is Epoch 101: 0.9679
Loss for Epoch 102: 0.1323641836643219
Accuracy for is Epoch 102: 0.9679
Loss for Epoch 103: 0.13083599507808685
Accuracy for is Epoch 103: 0.9679
Loss for Epoch 104: 0.12931756675243378
Accuracy for is Epoch 104: 0.9679
Loss for Epoch 105: 0.12780915200710297
Accuracy for is Epoch 105: 0.9808
Loss for Epoch 106: 0.12631098926067352
Accuracy for is Epoch 106: 0.9808
Loss for Epoch 107: 0.12482333928346634
Accuracy for is Epoch 107: 0.9808
Loss for Epoch 108: 0.12334650009870529
Accuracy for is Epoch 108: 0.9808
Loss for Epoch 109: 0.12188071757555008
Accuracy for is Epoch 109: 0.9808
Loss for Epoch 110: 0.1204262375831604
Accuracy for is Epoch 110: 0.9808
Loss for Epoch 111: 0.11898329108953476
Accuracy for is Epoch 111: 0.9808
Loss for Epoch 112: 0.11755212396383286
Accuracy for is Epoch 112: 0.9808
Loss for Epoch 113: 0.11613290756940842
Accuracy for is Epoch 113: 0.9808
Loss for Epoch 114: 0.11472588032484055
Accuracy for is Epoch 114: 0.9808
Loss for Epoch 115: 0.11333119869232178
Accuracy for is Epoch 115: 0.9808
Loss for Epoch 116: 0.11194900423288345
Accuracy for is Epoch 116: 0.9808
Loss for Epoch 117: 0.1105794906616211
Accuracy for is Epoch 117: 0.9808
Loss for Epoch 118: 0.10922273993492126
Accuracy for is Epoch 118: 0.9808
Loss for Epoch 119: 0.1078789085149765
Accuracy for is Epoch 119: 0.9808
Loss for Epoch 120: 0.10654806345701218
Accuracy for is Epoch 120: 0.9808
Loss for Epoch 121: 0.10523035377264023
Accuracy for is Epoch 121: 0.9808
Loss for Epoch 122: 0.10392585396766663
Accuracy for is Epoch 122: 0.9808
Loss for Epoch 123: 0.10263462364673615
Accuracy for is Epoch 123: 0.9808
Loss for Epoch 124: 0.10135678201913834
Accuracy for is Epoch 124: 0.9808
Loss for Epoch 125: 0.1000923365354538
Accuracy for is Epoch 125: 0.9808
Loss for Epoch 126: 0.09884139895439148
Accuracy for is Epoch 126: 0.9808
Loss for Epoch 127: 0.09760398417711258
Accuracy for is Epoch 127: 0.9808
Loss for Epoch 128: 0.09638015925884247
Accuracy for is Epoch 128: 0.9808
Loss for Epoch 129: 0.09516997635364532
Accuracy for is Epoch 129: 0.9808
Loss for Epoch 130: 0.09397348016500473
Accuracy for is Epoch 130: 0.9808
Loss for Epoch 131: 0.09279068559408188
Accuracy for is Epoch 131: 0.9808
Loss for Epoch 132: 0.09162164479494095
Accuracy for is Epoch 132: 0.9808
Loss for Epoch 133: 0.09046636521816254
Accuracy for is Epoch 133: 0.9808
Loss for Epoch 134: 0.08932486921548843
Accuracy for is Epoch 134: 0.9808
Loss for Epoch 135: 0.08819719403982162
Accuracy for is Epoch 135: 0.9808
Loss for Epoch 136: 0.0870833545923233
Accuracy for is Epoch 136: 0.9808
Loss for Epoch 137: 0.08598333597183228
Accuracy for is Epoch 137: 0.9808
Loss for Epoch 138: 0.08489716053009033
Accuracy for is Epoch 138: 0.9808
Loss for Epoch 139: 0.08382480591535568
Accuracy for is Epoch 139: 0.9808
Loss for Epoch 140: 0.08276629447937012
Accuracy for is Epoch 140: 0.9808
Loss for Epoch 141: 0.08172160387039185
Accuracy for is Epoch 141: 0.9808
Loss for Epoch 142: 0.08069069683551788
Accuracy for is Epoch 142: 0.9808
Loss for Epoch 143: 0.07967356592416763
Accuracy for is Epoch 143: 0.9808
Loss for Epoch 144: 0.07867017388343811
Accuracy for is Epoch 144: 0.9808
Loss for Epoch 145: 0.07768049836158752
Accuracy for is Epoch 145: 0.9808
Loss for Epoch 146: 0.0767044872045517
Accuracy for is Epoch 146: 0.9808
Loss for Epoch 147: 0.07574208825826645
Accuracy for is Epoch 147: 0.9872
Loss for Epoch 148: 0.0747932642698288
Accuracy for is Epoch 148: 0.9872
Loss for Epoch 149: 0.07385794073343277
Accuracy for is Epoch 149: 0.9872
Loss for Epoch 150: 0.07293606549501419
Accuracy for is Epoch 150: 0.9872
Loss for Epoch 151: 0.07202755659818649
Accuracy for is Epoch 151: 0.9872
Loss for Epoch 152: 0.07113233953714371
Accuracy for is Epoch 152: 0.9872
Loss for Epoch 153: 0.07025035470724106
Accuracy for is Epoch 153: 0.9872
Loss for Epoch 154: 0.069381482899189
Accuracy for is Epoch 154: 0.9936
Loss for Epoch 155: 0.06852562725543976
Accuracy for is Epoch 155: 0.9936
Loss for Epoch 156: 0.06768272072076797
Accuracy for is Epoch 156: 0.9936
Loss for Epoch 157: 0.0668526440858841
Accuracy for is Epoch 157: 0.9936
Loss for Epoch 158: 0.06603528559207916
Accuracy for is Epoch 158: 0.9936
Loss for Epoch 159: 0.06523054838180542
Accuracy for is Epoch 159: 0.9936
Loss for Epoch 160: 0.06443829834461212
Accuracy for is Epoch 160: 0.9936
Loss for Epoch 161: 0.06365841627120972
Accuracy for is Epoch 161: 0.9936
Loss for Epoch 162: 0.06289078295230865
Accuracy for is Epoch 162: 0.9936
Loss for Epoch 163: 0.06213526800274849
Accuracy for is Epoch 163: 0.9936
Loss for Epoch 164: 0.06139172613620758
Accuracy for is Epoch 164: 1.0
Loss for Epoch 165: 0.060660045593976974
Accuracy for is Epoch 165: 1.0
Loss for Epoch 166: 0.05994005128741264
Accuracy for is Epoch 166: 1.0
Loss for Epoch 167: 0.05923163518309593
Accuracy for is Epoch 167: 1.0
Loss for Epoch 168: 0.058534640818834305
Accuracy for is Epoch 168: 1.0
Loss for Epoch 169: 0.05784890428185463
Accuracy for is Epoch 169: 1.0
Loss for Epoch 170: 0.05717430263757706
Accuracy for is Epoch 170: 1.0
Loss for Epoch 171: 0.05651067569851875
Accuracy for is Epoch 171: 1.0
Loss for Epoch 172: 0.05585784092545509
Accuracy for is Epoch 172: 1.0
Loss for Epoch 173: 0.05521570146083832
Accuracy for is Epoch 173: 1.0
Loss for Epoch 174: 0.05458405241370201
Accuracy for is Epoch 174: 1.0
Loss for Epoch 175: 0.05396275594830513
Accuracy for is Epoch 175: 1.0
Loss for Epoch 176: 0.05335165932774544
Accuracy for is Epoch 176: 1.0
Loss for Epoch 177: 0.0527505949139595
Accuracy for is Epoch 177: 1.0
Loss for Epoch 178: 0.05215940624475479
Accuracy for is Epoch 178: 1.0
Loss for Epoch 179: 0.051577940583229065
Accuracy for is Epoch 179: 1.0
Loss for Epoch 180: 0.05100603401660919
Accuracy for is Epoch 180: 1.0
Loss for Epoch 181: 0.050443537533283234
Accuracy for is Epoch 181: 1.0
Loss for Epoch 182: 0.04989028349518776
Accuracy for is Epoch 182: 1.0
Loss for Epoch 183: 0.049346115440130234
Accuracy for is Epoch 183: 1.0
Loss for Epoch 184: 0.04881088435649872
Accuracy for is Epoch 184: 1.0
Loss for Epoch 185: 0.048284441232681274
Accuracy for is Epoch 185: 1.0
Loss for Epoch 186: 0.04776662588119507
Accuracy for is Epoch 186: 1.0
Loss for Epoch 187: 0.04725727438926697
Accuracy for is Epoch 187: 1.0
Loss for Epoch 188: 0.046756256371736526
Accuracy for is Epoch 188: 1.0
Loss for Epoch 189: 0.04626340791583061
Accuracy for is Epoch 189: 1.0
Loss for Epoch 190: 0.045778583735227585
Accuracy for is Epoch 190: 1.0
Loss for Epoch 191: 0.0453016422688961
Accuracy for is Epoch 191: 1.0
Loss for Epoch 192: 0.044832438230514526
Accuracy for is Epoch 192: 1.0
Loss for Epoch 193: 0.04437081888318062
Accuracy for is Epoch 193: 1.0
Loss for Epoch 194: 0.04391665384173393
Accuracy for is Epoch 194: 1.0
Loss for Epoch 195: 0.04346981272101402
Accuracy for is Epoch 195: 1.0
Loss for Epoch 196: 0.04303012788295746
Accuracy for is Epoch 196: 1.0
Loss for Epoch 197: 0.04259749501943588
Accuracy for is Epoch 197: 1.0
Loss for Epoch 198: 0.04217176139354706
Accuracy for is Epoch 198: 1.0
Loss for Epoch 199: 0.041752807796001434
Accuracy for is Epoch 199: 1.0
Loss for Epoch 200: 0.04134050011634827
Accuracy for is Epoch 200: 1.0
Loss for Epoch 201: 0.040934719145298004
Accuracy for is Epoch 201: 1.0
Loss for Epoch 202: 0.040535323321819305
Accuracy for is Epoch 202: 1.0
Loss for Epoch 203: 0.040142204612493515
Accuracy for is Epoch 203: 1.0
Loss for Epoch 204: 0.03975524380803108
Accuracy for is Epoch 204: 1.0
Loss for Epoch 205: 0.03937431424856186
Accuracy for is Epoch 205: 1.0
Loss for Epoch 206: 0.038999300450086594
Accuracy for is Epoch 206: 1.0
Loss for Epoch 207: 0.03863010182976723
Accuracy for is Epoch 207: 1.0
Loss for Epoch 208: 0.03826659172773361
Accuracy for is Epoch 208: 1.0
Loss for Epoch 209: 0.037908658385276794
Accuracy for is Epoch 209: 1.0
Loss for Epoch 210: 0.03755621239542961
Accuracy for is Epoch 210: 1.0
Loss for Epoch 211: 0.03720913827419281
Accuracy for is Epoch 211: 1.0
Loss for Epoch 212: 0.036867327988147736
Accuracy for is Epoch 212: 1.0
Loss for Epoch 213: 0.036530692130327225
Accuracy for is Epoch 213: 1.0
Loss for Epoch 214: 0.03619911149144173
Accuracy for is Epoch 214: 1.0
Loss for Epoch 215: 0.03587250038981438
Accuracy for is Epoch 215: 1.0
Loss for Epoch 216: 0.03555077686905861
Accuracy for is Epoch 216: 1.0
Loss for Epoch 217: 0.03523382544517517
Accuracy for is Epoch 217: 1.0
Loss for Epoch 218: 0.03492157161235809
Accuracy for is Epoch 218: 1.0
Loss for Epoch 219: 0.03461391478776932
Accuracy for is Epoch 219: 1.0
Loss for Epoch 220: 0.034310776740312576
Accuracy for is Epoch 220: 1.0
Loss for Epoch 221: 0.03401205316185951
Accuracy for is Epoch 221: 1.0
Loss for Epoch 222: 0.03371768817305565
Accuracy for is Epoch 222: 1.0
Loss for Epoch 223: 0.03342757374048233
Accuracy for is Epoch 223: 1.0
Loss for Epoch 224: 0.033141642808914185
Accuracy for is Epoch 224: 1.0
Loss for Epoch 225: 0.03285980969667435
Accuracy for is Epoch 225: 1.0
Loss for Epoch 226: 0.032582007348537445
Accuracy for is Epoch 226: 1.0
Loss for Epoch 227: 0.03230816125869751
Accuracy for is Epoch 227: 1.0
Loss for Epoch 228: 0.032038185745477676
Accuracy for is Epoch 228: 1.0
Loss for Epoch 229: 0.031772010028362274
Accuracy for is Epoch 229: 1.0
Loss for Epoch 230: 0.03150956332683563
Accuracy for is Epoch 230: 1.0
Loss for Epoch 231: 0.03125078231096268
Accuracy for is Epoch 231: 1.0
Loss for Epoch 232: 0.030995601788163185
Accuracy for is Epoch 232: 1.0
Loss for Epoch 233: 0.030743947252631187
Accuracy for is Epoch 233: 1.0
Loss for Epoch 234: 0.030495749786496162
Accuracy for is Epoch 234: 1.0
Loss for Epoch 235: 0.030250955373048782
Accuracy for is Epoch 235: 1.0
Loss for Epoch 236: 0.030009504407644272
Accuracy for is Epoch 236: 1.0
Loss for Epoch 237: 0.02977132797241211
Accuracy for is Epoch 237: 1.0
Loss for Epoch 238: 0.029536370187997818
Accuracy for is Epoch 238: 1.0
Loss for Epoch 239: 0.02930457517504692
Accuracy for is Epoch 239: 1.0
Loss for Epoch 240: 0.029075879603624344
Accuracy for is Epoch 240: 1.0
Loss for Epoch 241: 0.02885022573173046
Accuracy for is Epoch 241: 1.0
Loss for Epoch 242: 0.02862757071852684
Accuracy for is Epoch 242: 1.0
Loss for Epoch 243: 0.028407849371433258
Accuracy for is Epoch 243: 1.0
Loss for Epoch 244: 0.028191013261675835
Accuracy for is Epoch 244: 1.0
Loss for Epoch 245: 0.02797701768577099
Accuracy for is Epoch 245: 1.0
Loss for Epoch 246: 0.027765803039073944
Accuracy for is Epoch 246: 1.0
Loss for Epoch 247: 0.027557330206036568
Accuracy for is Epoch 247: 1.0
Loss for Epoch 248: 0.027351537719368935
Accuracy for is Epoch 248: 1.0
Loss for Epoch 249: 0.027148384600877762
Accuracy for is Epoch 249: 1.0
Loss for Epoch 250: 0.026947828009724617
Accuracy for is Epoch 250: 1.0
Loss for Epoch 251: 0.026749828830361366
Accuracy for is Epoch 251: 1.0
Loss for Epoch 252: 0.026554325595498085
Accuracy for is Epoch 252: 1.0
Loss for Epoch 253: 0.026361290365457535
Accuracy for is Epoch 253: 1.0
Loss for Epoch 254: 0.026170672848820686
Accuracy for is Epoch 254: 1.0
Loss for Epoch 255: 0.025982435792684555
Accuracy for is Epoch 255: 1.0
Loss for Epoch 256: 0.025796538218855858
Accuracy for is Epoch 256: 1.0
Loss for Epoch 257: 0.025612935423851013
Accuracy for is Epoch 257: 1.0
Loss for Epoch 258: 0.025431593880057335
Accuracy for is Epoch 258: 1.0
Loss for Epoch 259: 0.02525247447192669
Accuracy for is Epoch 259: 1.0
Loss for Epoch 260: 0.025075532495975494
Accuracy for is Epoch 260: 1.0
Loss for Epoch 261: 0.02490074560046196
Accuracy for is Epoch 261: 1.0
Loss for Epoch 262: 0.024728069081902504
Accuracy for is Epoch 262: 1.0
Loss for Epoch 263: 0.024557465687394142
Accuracy for is Epoch 263: 1.0
Loss for Epoch 264: 0.024388913065195084
Accuracy for is Epoch 264: 1.0
Loss for Epoch 265: 0.024222364649176598
Accuracy for is Epoch 265: 1.0
Loss for Epoch 266: 0.024057796224951744
Accuracy for is Epoch 266: 1.0
Loss for Epoch 267: 0.023895159363746643
Accuracy for is Epoch 267: 1.0
Loss for Epoch 268: 0.0237344391644001
Accuracy for is Epoch 268: 1.0
Loss for Epoch 269: 0.023575609549880028
Accuracy for is Epoch 269: 1.0
Loss for Epoch 270: 0.023418625816702843
Accuracy for is Epoch 270: 1.0
Loss for Epoch 271: 0.023263458162546158
Accuracy for is Epoch 271: 1.0
Loss for Epoch 272: 0.023110080510377884
Accuracy for is Epoch 272: 1.0
Loss for Epoch 273: 0.02295847237110138
Accuracy for is Epoch 273: 1.0
Loss for Epoch 274: 0.022808605805039406
Accuracy for is Epoch 274: 1.0
Loss for Epoch 275: 0.022660430520772934
Accuracy for is Epoch 275: 1.0
Loss for Epoch 276: 0.022513946518301964
Accuracy for is Epoch 276: 1.0
Loss for Epoch 277: 0.02236912027001381
Accuracy for is Epoch 277: 1.0
Loss for Epoch 278: 0.022225912660360336
Accuracy for is Epoch 278: 1.0
Loss for Epoch 279: 0.022084318101406097
Accuracy for is Epoch 279: 1.0
Loss for Epoch 280: 0.021944290027022362
Accuracy for is Epoch 280: 1.0
Loss for Epoch 281: 0.02180582284927368
Accuracy for is Epoch 281: 1.0
Loss for Epoch 282: 0.02166888304054737
Accuracy for is Epoch 282: 1.0
Loss for Epoch 283: 0.021533453837037086
Accuracy for is Epoch 283: 1.0
Loss for Epoch 284: 0.021399501711130142
Accuracy for is Epoch 284: 1.0
Loss for Epoch 285: 0.021267011761665344
Accuracy for is Epoch 285: 1.0
Loss for Epoch 286: 0.021135959774255753
Accuracy for is Epoch 286: 1.0
Loss for Epoch 287: 0.021006332710385323
Accuracy for is Epoch 287: 1.0
Loss for Epoch 288: 0.02087809145450592
Accuracy for is Epoch 288: 1.0
Loss for Epoch 289: 0.0207512266933918
Accuracy for is Epoch 289: 1.0
Loss for Epoch 290: 0.02062571607530117
Accuracy for is Epoch 290: 1.0
Loss for Epoch 291: 0.02050154097378254
Accuracy for is Epoch 291: 1.0
Loss for Epoch 292: 0.020378679037094116
Accuracy for is Epoch 292: 1.0
Loss for Epoch 293: 0.020257113501429558
Accuracy for is Epoch 293: 1.0
Loss for Epoch 294: 0.02013682946562767
Accuracy for is Epoch 294: 1.0
Loss for Epoch 295: 0.02001778967678547
Accuracy for is Epoch 295: 1.0
Loss for Epoch 296: 0.019899997860193253
Accuracy for is Epoch 296: 1.0
Loss for Epoch 297: 0.019783422350883484
Accuracy for is Epoch 297: 1.0
Loss for Epoch 298: 0.019668051972985268
Accuracy for is Epoch 298: 1.0
Loss for Epoch 299: 0.019553856924176216
Accuracy for is Epoch 299: 1.0
Loss for Epoch 300: 0.01944083720445633
Accuracy for is Epoch 300: 1.0
Loss for Epoch 301: 0.019328976050019264
Accuracy for is Epoch 301: 1.0
Loss for Epoch 302: 0.01921824924647808
Accuracy for is Epoch 302: 1.0
Loss for Epoch 303: 0.019108640030026436
Accuracy for is Epoch 303: 1.0
Loss for Epoch 304: 0.01900012418627739
Accuracy for is Epoch 304: 1.0
Loss for Epoch 305: 0.01889270916581154
Accuracy for is Epoch 305: 1.0
Loss for Epoch 306: 0.018786363303661346
Accuracy for is Epoch 306: 1.0
Loss for Epoch 307: 0.018681077286601067
Accuracy for is Epoch 307: 1.0
Loss for Epoch 308: 0.01857682131230831
Accuracy for is Epoch 308: 1.0
Loss for Epoch 309: 0.01847360096871853
Accuracy for is Epoch 309: 1.0
Loss for Epoch 310: 0.018371401354670525
Accuracy for is Epoch 310: 1.0
Loss for Epoch 311: 0.01827019825577736
Accuracy for is Epoch 311: 1.0
Loss for Epoch 312: 0.018169982358813286
Accuracy for is Epoch 312: 1.0
Loss for Epoch 313: 0.018070736899971962
Accuracy for is Epoch 313: 1.0
Loss for Epoch 314: 0.017972448840737343
Accuracy for is Epoch 314: 1.0
Loss for Epoch 315: 0.017875107005238533
Accuracy for is Epoch 315: 1.0
Loss for Epoch 316: 0.017778705805540085
Accuracy for is Epoch 316: 1.0
Loss for Epoch 317: 0.01768322102725506
Accuracy for is Epoch 317: 1.0
Loss for Epoch 318: 0.017588647082448006
Accuracy for is Epoch 318: 1.0
Loss for Epoch 319: 0.01749497279524803
Accuracy for is Epoch 319: 1.0
Loss for Epoch 320: 0.017402177676558495
Accuracy for is Epoch 320: 1.0
Loss for Epoch 321: 0.017310254275798798
Accuracy for is Epoch 321: 1.0
Loss for Epoch 322: 0.017219197005033493
Accuracy for is Epoch 322: 1.0
Loss for Epoch 323: 0.01712898537516594
Accuracy for is Epoch 323: 1.0
Loss for Epoch 324: 0.017039615660905838
Accuracy for is Epoch 324: 1.0
Loss for Epoch 325: 0.016951076686382294
Accuracy for is Epoch 325: 1.0
Loss for Epoch 326: 0.016863351687788963
Accuracy for is Epoch 326: 1.0
Loss for Epoch 327: 0.0167764313519001
Accuracy for is Epoch 327: 1.0
Loss for Epoch 328: 0.016690311953425407
Accuracy for is Epoch 328: 1.0
Loss for Epoch 329: 0.01660497486591339
Accuracy for is Epoch 329: 1.0
Loss for Epoch 330: 0.016520414501428604
Accuracy for is Epoch 330: 1.0
Loss for Epoch 331: 0.016436615958809853
Accuracy for is Epoch 331: 1.0
Loss for Epoch 332: 0.016353577375411987
Accuracy for is Epoch 332: 1.0
Loss for Epoch 333: 0.016271281987428665
Accuracy for is Epoch 333: 1.0
Loss for Epoch 334: 0.016189727932214737
Accuracy for is Epoch 334: 1.0
Loss for Epoch 335: 0.01610890030860901
Accuracy for is Epoch 335: 1.0
Loss for Epoch 336: 0.016028789803385735
Accuracy for is Epoch 336: 1.0
Loss for Epoch 337: 0.015949388965964317
Accuracy for is Epoch 337: 1.0
Loss for Epoch 338: 0.01587069034576416
Accuracy for is Epoch 338: 1.0
Loss for Epoch 339: 0.015792682766914368
Accuracy for is Epoch 339: 1.0
Loss for Epoch 340: 0.015715356916189194
Accuracy for is Epoch 340: 1.0
Loss for Epoch 341: 0.01563870720565319
Accuracy for is Epoch 341: 1.0
Loss for Epoch 342: 0.01556272804737091
Accuracy for is Epoch 342: 1.0
Loss for Epoch 343: 0.01548740454018116
Accuracy for is Epoch 343: 1.0
Loss for Epoch 344: 0.015412723645567894
Accuracy for is Epoch 344: 1.0
Loss for Epoch 345: 0.01533869095146656
Accuracy for is Epoch 345: 1.0
Loss for Epoch 346: 0.015265293419361115
Accuracy for is Epoch 346: 1.0
Loss for Epoch 347: 0.015192518010735512
Accuracy for is Epoch 347: 1.0
Loss for Epoch 348: 0.015120365656912327
Accuracy for is Epoch 348: 1.0
Loss for Epoch 349: 0.015048825182020664
Accuracy for is Epoch 349: 1.0
Loss for Epoch 350: 0.014977889135479927
Accuracy for is Epoch 350: 1.0
Loss for Epoch 351: 0.014907552860677242
Accuracy for is Epoch 351: 1.0
Loss for Epoch 352: 0.014837798662483692
Accuracy for is Epoch 352: 1.0
Loss for Epoch 353: 0.014768628403544426
Accuracy for is Epoch 353: 1.0
Loss for Epoch 354: 0.014700040221214294
Accuracy for is Epoch 354: 1.0
Loss for Epoch 355: 0.014632027596235275
Accuracy for is Epoch 355: 1.0
Loss for Epoch 356: 0.014564563520252705
Accuracy for is Epoch 356: 1.0
Loss for Epoch 357: 0.014497663825750351
Accuracy for is Epoch 357: 1.0
Loss for Epoch 358: 0.014431305229663849
Accuracy for is Epoch 358: 1.0
Loss for Epoch 359: 0.014365500770509243
Accuracy for is Epoch 359: 1.0
Loss for Epoch 360: 0.014300225302577019
Accuracy for is Epoch 360: 1.0
Loss for Epoch 361: 0.01423549372702837
Accuracy for is Epoch 361: 1.0
Loss for Epoch 362: 0.014171268790960312
Accuracy for is Epoch 362: 1.0
Loss for Epoch 363: 0.014107570052146912
Accuracy for is Epoch 363: 1.0
Loss for Epoch 364: 0.014044385403394699
Accuracy for is Epoch 364: 1.0
Loss for Epoch 365: 0.013981706462800503
Accuracy for is Epoch 365: 1.0
Loss for Epoch 366: 0.013919529505074024
Accuracy for is Epoch 366: 1.0
Loss for Epoch 367: 0.013857843354344368
Accuracy for is Epoch 367: 1.0
Loss for Epoch 368: 0.013796651735901833
Accuracy for is Epoch 368: 1.0
Loss for Epoch 369: 0.013735939748585224
Accuracy for is Epoch 369: 1.0
Loss for Epoch 370: 0.013675708323717117
Accuracy for is Epoch 370: 1.0
Loss for Epoch 371: 0.013615953736007214
Accuracy for is Epoch 371: 1.0
Loss for Epoch 372: 0.013556666672229767
Accuracy for is Epoch 372: 1.0
Loss for Epoch 373: 0.013497831299901009
Accuracy for is Epoch 373: 1.0
Loss for Epoch 374: 0.01343946997076273
Accuracy for is Epoch 374: 1.0
Loss for Epoch 375: 0.013381557539105415
Accuracy for is Epoch 375: 1.0
Loss for Epoch 376: 0.013324081897735596
Accuracy for is Epoch 376: 1.0
Loss for Epoch 377: 0.013267059810459614
Accuracy for is Epoch 377: 1.0
Loss for Epoch 378: 0.013210475444793701
Accuracy for is Epoch 378: 1.0
Loss for Epoch 379: 0.013154320418834686
Accuracy for is Epoch 379: 1.0
Loss for Epoch 380: 0.01309859286993742
Accuracy for is Epoch 380: 1.0
Loss for Epoch 381: 0.013043294660747051
Accuracy for is Epoch 381: 1.0
Loss for Epoch 382: 0.012988414615392685
Accuracy for is Epoch 382: 1.0
Loss for Epoch 383: 0.012933946214616299
Accuracy for is Epoch 383: 1.0
Loss for Epoch 384: 0.012879888527095318
Accuracy for is Epoch 384: 1.0
Loss for Epoch 385: 0.012826235964894295
Accuracy for is Epoch 385: 1.0
Loss for Epoch 386: 0.012772982008755207
Accuracy for is Epoch 386: 1.0
Loss for Epoch 387: 0.012720133177936077
Accuracy for is Epoch 387: 1.0
Loss for Epoch 388: 0.012667671777307987
Accuracy for is Epoch 388: 1.0
Loss for Epoch 389: 0.01261560432612896
Accuracy for is Epoch 389: 1.0
Loss for Epoch 390: 0.0125639159232378
Accuracy for is Epoch 390: 1.0
Loss for Epoch 391: 0.012512610293924809
Accuracy for is Epoch 391: 1.0
Loss for Epoch 392: 0.012461683712899685
Accuracy for is Epoch 392: 1.0
Loss for Epoch 393: 0.012411129660904408
Accuracy for is Epoch 393: 1.0
Loss for Epoch 394: 0.012360942550003529
Accuracy for is Epoch 394: 1.0
Loss for Epoch 395: 0.012311122380197048
Accuracy for is Epoch 395: 1.0
Loss for Epoch 396: 0.012261662632226944
Accuracy for is Epoch 396: 1.0
Loss for Epoch 397: 0.012212551198899746
Accuracy for is Epoch 397: 1.0
Loss for Epoch 398: 0.012163807637989521
Accuracy for is Epoch 398: 1.0
Loss for Epoch 399: 0.012115409597754478
Accuracy for is Epoch 399: 1.0
Loss for Epoch 400: 0.012067358940839767
Accuracy for is Epoch 400: 1.0
Loss for Epoch 401: 0.012019647285342216
Accuracy for is Epoch 401: 1.0
Loss for Epoch 402: 0.011972276493906975
Accuracy for is Epoch 402: 1.0
Loss for Epoch 403: 0.011925249360501766
Accuracy for is Epoch 403: 1.0
Loss for Epoch 404: 0.011878546327352524
Accuracy for is Epoch 404: 1.0
Loss for Epoch 405: 0.011832175776362419
Accuracy for is Epoch 405: 1.0
Loss for Epoch 406: 0.011786133982241154
Accuracy for is Epoch 406: 1.0
Loss for Epoch 407: 0.011740412563085556
Accuracy for is Epoch 407: 1.0
Loss for Epoch 408: 0.011695006862282753
Accuracy for is Epoch 408: 1.0
Loss for Epoch 409: 0.011649922467768192
Accuracy for is Epoch 409: 1.0
Loss for Epoch 410: 0.011605150066316128
Accuracy for is Epoch 410: 1.0
Loss for Epoch 411: 0.011560685001313686
Accuracy for is Epoch 411: 1.0
Loss for Epoch 412: 0.011516541242599487
Accuracy for is Epoch 412: 1.0
Loss for Epoch 413: 0.01147268433123827
Accuracy for is Epoch 413: 1.0
Loss for Epoch 414: 0.0114291375502944
Accuracy for is Epoch 414: 1.0
Loss for Epoch 415: 0.011385884135961533
Accuracy for is Epoch 415: 1.0
Loss for Epoch 416: 0.011342927813529968
Accuracy for is Epoch 416: 1.0
Loss for Epoch 417: 0.011300265789031982
Accuracy for is Epoch 417: 1.0
Loss for Epoch 418: 0.011257889680564404
Accuracy for is Epoch 418: 1.0
Loss for Epoch 419: 0.011215806007385254
Accuracy for is Epoch 419: 1.0
Loss for Epoch 420: 0.011174005456268787
Accuracy for is Epoch 420: 1.0
Loss for Epoch 421: 0.011132482439279556
Accuracy for is Epoch 421: 1.0
Loss for Epoch 422: 0.01109123695641756
Accuracy for is Epoch 422: 1.0
Loss for Epoch 423: 0.01105027087032795
Accuracy for is Epoch 423: 1.0
Loss for Epoch 424: 0.011009575799107552
Accuracy for is Epoch 424: 1.0
Loss for Epoch 425: 0.010969155468046665
Accuracy for is Epoch 425: 1.0
Loss for Epoch 426: 0.010928994044661522
Accuracy for is Epoch 426: 1.0
Loss for Epoch 427: 0.010889106430113316
Accuracy for is Epoch 427: 1.0
Loss for Epoch 428: 0.010849477723240852
Accuracy for is Epoch 428: 1.0
Loss for Epoch 429: 0.010810110718011856
Accuracy for is Epoch 429: 1.0
Loss for Epoch 430: 0.010771005414426327
Accuracy for is Epoch 430: 1.0
Loss for Epoch 431: 0.010732153430581093
Accuracy for is Epoch 431: 1.0
Loss for Epoch 432: 0.01069355383515358
Accuracy for is Epoch 432: 1.0
Loss for Epoch 433: 0.010655202902853489
Accuracy for is Epoch 433: 1.0
Loss for Epoch 434: 0.010617105290293694
Accuracy for is Epoch 434: 1.0
Loss for Epoch 435: 0.010579249821603298
Accuracy for is Epoch 435: 1.0
Loss for Epoch 436: 0.010541647672653198
Accuracy for is Epoch 436: 1.0
Loss for Epoch 437: 0.010504275560379028
Accuracy for is Epoch 437: 1.0
Loss for Epoch 438: 0.010467151179909706
Accuracy for is Epoch 438: 1.0
Loss for Epoch 439: 0.01043025404214859
Accuracy for is Epoch 439: 1.0
Loss for Epoch 440: 0.010393607430160046
Accuracy for is Epoch 440: 1.0
Loss for Epoch 441: 0.010357179678976536
Accuracy for is Epoch 441: 1.0
Loss for Epoch 442: 0.010320993140339851
Accuracy for is Epoch 442: 1.0
Loss for Epoch 443: 0.010285023599863052
Accuracy for is Epoch 443: 1.0
Loss for Epoch 444: 0.010249285958707333
Accuracy for is Epoch 444: 1.0
Loss for Epoch 445: 0.010213782079517841
Accuracy for is Epoch 445: 1.0
Loss for Epoch 446: 0.010178486816585064
Accuracy for is Epoch 446: 1.0
Loss for Epoch 447: 0.01014342624694109
Accuracy for is Epoch 447: 1.0
Loss for Epoch 448: 0.010108571499586105
Accuracy for is Epoch 448: 1.0
Loss for Epoch 449: 0.010073940269649029
Accuracy for is Epoch 449: 1.0
Loss for Epoch 450: 0.010039526969194412
Accuracy for is Epoch 450: 1.0
Loss for Epoch 451: 0.010005323216319084
Accuracy for is Epoch 451: 1.0
Loss for Epoch 452: 0.00997132807970047
Accuracy for is Epoch 452: 1.0
Loss for Epoch 453: 0.009937545284628868
Accuracy for is Epoch 453: 1.0
Loss for Epoch 454: 0.009903961792588234
Accuracy for is Epoch 454: 1.0
Loss for Epoch 455: 0.009870597161352634
Accuracy for is Epoch 455: 1.0
Loss for Epoch 456: 0.009837429039180279
Accuracy for is Epoch 456: 1.0
Loss for Epoch 457: 0.00980446208268404
Accuracy for is Epoch 457: 1.0
Loss for Epoch 458: 0.009771698154509068
Accuracy for is Epoch 458: 1.0
Loss for Epoch 459: 0.009739132598042488
Accuracy for is Epoch 459: 1.0
Loss for Epoch 460: 0.009706763550639153
Accuracy for is Epoch 460: 1.0
Loss for Epoch 461: 0.009674584493041039
Accuracy for is Epoch 461: 1.0
Loss for Epoch 462: 0.009642601013183594
Accuracy for is Epoch 462: 1.0
Loss for Epoch 463: 0.009610812179744244
Accuracy for is Epoch 463: 1.0
Loss for Epoch 464: 0.009579211473464966
Accuracy for is Epoch 464: 1.0
Loss for Epoch 465: 0.009547803550958633
Accuracy for is Epoch 465: 1.0
Loss for Epoch 466: 0.009516575373709202
Accuracy for is Epoch 466: 1.0
Loss for Epoch 467: 0.009485539048910141
Accuracy for is Epoch 467: 1.0
Loss for Epoch 468: 0.009454679675400257
Accuracy for is Epoch 468: 1.0
Loss for Epoch 469: 0.009424007497727871
Accuracy for is Epoch 469: 1.0
Loss for Epoch 470: 0.009393512271344662
Accuracy for is Epoch 470: 1.0
Loss for Epoch 471: 0.009363200515508652
Accuracy for is Epoch 471: 1.0
Loss for Epoch 472: 0.009333064779639244
Accuracy for is Epoch 472: 1.0
Loss for Epoch 473: 0.009303102269768715
Accuracy for is Epoch 473: 1.0
Loss for Epoch 474: 0.009273318573832512
Accuracy for is Epoch 474: 1.0
Loss for Epoch 475: 0.009243707172572613
Accuracy for is Epoch 475: 1.0
Loss for Epoch 476: 0.009214267134666443
Accuracy for is Epoch 476: 1.0
Loss for Epoch 477: 0.009184992872178555
Accuracy for is Epoch 477: 1.0
Loss for Epoch 478: 0.00915589090436697
Accuracy for is Epoch 478: 1.0
Loss for Epoch 479: 0.009126957505941391
Accuracy for is Epoch 479: 1.0
Loss for Epoch 480: 0.009098188020288944
Accuracy for is Epoch 480: 1.0
Loss for Epoch 481: 0.009069581516087055
Accuracy for is Epoch 481: 1.0
Loss for Epoch 482: 0.009041142649948597
Accuracy for is Epoch 482: 1.0
Loss for Epoch 483: 0.009012863039970398
Accuracy for is Epoch 483: 1.0
Loss for Epoch 484: 0.008984743617475033
Accuracy for is Epoch 484: 1.0
Loss for Epoch 485: 0.008956785313785076
Accuracy for is Epoch 485: 1.0
Loss for Epoch 486: 0.008928988128900528
Accuracy for is Epoch 486: 1.0
Loss for Epoch 487: 0.008901339955627918
Accuracy for is Epoch 487: 1.0
Loss for Epoch 488: 0.008873851969838142
Accuracy for is Epoch 488: 1.0
Loss for Epoch 489: 0.008846525102853775
Accuracy for is Epoch 489: 1.0
Loss for Epoch 490: 0.00881933979690075
Accuracy for is Epoch 490: 1.0
Loss for Epoch 491: 0.008792308159172535
Accuracy for is Epoch 491: 1.0
Loss for Epoch 492: 0.008765429258346558
Accuracy for is Epoch 492: 1.0
Loss for Epoch 493: 0.008738703094422817
Accuracy for is Epoch 493: 1.0
Loss for Epoch 494: 0.008712121285498142
Accuracy for is Epoch 494: 1.0
Loss for Epoch 495: 0.008685688488185406
Accuracy for is Epoch 495: 1.0
Loss for Epoch 496: 0.00865940097719431
Accuracy for is Epoch 496: 1.0
Loss for Epoch 497: 0.008633255027234554
Accuracy for is Epoch 497: 1.0
Loss for Epoch 498: 0.008607256226241589
Accuracy for is Epoch 498: 1.0
Loss for Epoch 499: 0.00858139805495739
Accuracy for is Epoch 499: 1.0
###Markdown
Visualize the final node embeddingsVisualize your final embedding here! You can visually compare the figure with the previous embedding figure. After training, you should oberserve that the two classes are more evidently separated. This is a great sanitity check for your implementation as well.
###Code
# Visualize the final learned embedding
visualize_emb(emb)
###Output
_____no_output_____ |
bumphunter/Python_notebook_stepping_through_bumphunter/py_bumphunter.ipynb | ###Markdown
Python-equivalent commands for bumphunting
###Code
#import csv ##module required so far just to read CSV data files output by minfi processing in R
import pandas as pd
import numpy as np
import statsmodels.api as sm
##Read in the minfi processed output matrices needed to run bumphunter code
betaVals = pd.read_csv("Sample_Beta_value_matrix_quantilePreprocessed.csv",index_col=0)
phenoDesign = pd.read_csv("Sample_phenotype_design_matrix.csv")
probePositions = pd.read_csv("Sample_probe_genomic_positions.csv",index_col=0)
probePositions.head()
betaVals.head()
betaVals.shape
###Output
_____no_output_____
###Markdown
The first set of commands within the bumphunterEngine check which arguments were specified when the bumphunter function was called, check if they're in the correct format, and produce informative warnings if any input is incorrect.Then the bumphunterEngine calls several functions from R's foreach package to see if multiple cores were registered before bumphunter was called. This sets up the makes it possible for the more time-consuming bumphunterEngine functions to process in parallel as coded.Parallel processing has NOT been addressed yet in this python implementationThe first backend function bumphunterEngine calls is clusterMaker, which assigns genomic locations to sets of clusters if they are within maxGap of one another (if clusters were not already pre-assigned in one of the function arguments).
###Code
##if (is.null(cluster))
##cluster <- clusterMaker(chr, pos, maxGap = maxGap)
###Output
_____no_output_____
###Markdown
clusterMaker() commands
###Code
##clusterMaker function args (lists a different default maxGap than bumphunter())
##clusterMaker <- function(chr, pos, assumeSorted = FALSE, maxGap=300)
#nonaIndex <- which(!is.na(chr) & !is.na(pos)) ##integer vector of probe indices from 1-485512
#Indexes <- split(nonaIndex, chr[nonaIndex]) ##24 integer vectors (one for each chromosome)
#clusterIDs <- rep(NA, length(chr))
#LAST <- 0
##Set maximum distance in bp allowed between probes in the same cluster
maxGap = 300
assumeSorted = False ##don't assume the genomic positions are listed in order of chromosomal location
##Drop NA values from the dataframe of chromosome IDs and genomic locations for each probe
if probePositions.isnull().values.any():
probePositions.dropna()
##set "last" counter to zero (keeps count of how many clusters were assigned on previous chromosomes in following loop)
last = 0
##Create list of all chromosome IDs present in dataset
probeChr = set(probePositions["Chromosome"])
probeChr = list(probeChr)
probeChr[:5]
##list to hold cluster ID number for each genomic position
clusterIDs = []
###Output
_____no_output_____
###Markdown
R code (commented out above Python commands) splits the probe positions by chromosome into a list of vectors, but is that really necessary? I'm going to keep the pandas dataframe together and select by chromosome when it comes to it. Might change this if that's unwieldy or slow.
###Code
#assumeSorted = FALSE
#for(i in seq(along = Indexes)){ ##loops 24 times, once for each chromosome
#Index <- Indexes[[i]]
#x <- pos[Index] ##select probe positions for that chromosome
#if(!assumeSorted){ ##make sure probe genomic locations are in order
#Index <- Index[order(x)]
#x <- pos[Index]
#}
##diff(x) to calculate distance between each sequential position
#y <- as.numeric(diff(x) > maxGap) ##binary vector where 0 = distance < maxGap and 1 = distance < maxGap
#z <- cumsum(c(1, y)) ##cumulative sum of this binary vector
#clusterIDs[Index] <- z + LAST ##turn this cumulative sum vector into the cluster ID numbers
#LAST <- max(z) + LAST ##record the last clusterID to start at the next number for the following chromosome
#}
#head(clusterIDs)
##Loop 24 times (once for each chromosome since clusters can't span across chromosomes)
for i in probeChr:
##Select just the rows of the dataframe corresponding to that chromosome
chr_rows = probePositions[probePositions.Chromosome==i]
##Create a series of the numeric genomic locations for each probe on that chromosome
x = chr_rows.Position
##Make sure they're sorted in ascending order
if not assumeSorted:
x = x.sort_values()
##Calculate the difference between each successive position
diff_x = np.diff(x)
##Create a list of 0s (difference is <maxGap) and 1s (difference is >maxGap)
y = (diff_x > maxGap)*1
##diff is always one item shorter than the series from which it was calculated, so add one 1 to the start of y
#y.insert(0,1)
y_all = np.insert(y,0,1)
##Take the cumulative sum of these zeroes and ones to assign cluster ID numbers
##that start a new cluster each time the probes are >maxGap apart
z = np.cumsum(y_all)
##Add the final cluster number from the previous chromosome to z so that this chromosome starts at the next ID #
#z_plus = [1+i for i in z]
z_plus = z+last
##Concatenate this chromosome's list of cluster IDs with the running list of cluster ID numbers
clusterIDs = np.append(clusterIDs,z_plus)
##Adjust the "last" counter to the final cluster ID number for this chromosome
last = max(z_plus)
len(clusterIDs)
clusterIDs[-10:]
###Output
_____no_output_____
###Markdown
After designating cluster ID numbers, bumphunterEngine makes decisions about smoothing and weighting the data depending on the inputs the user provided for the function. Whether the data is eventually weighted or not, it is first processed through the .getEstimate method.
###Code
##if (useWeights & smooth) {
##tmp <- .getEstimate(mat = mat, design = design,
##coef = coef, full = TRUE)
##rawBeta <- tmp$coef
##weights <- 1/tmp$sigma
##rm(tmp)
##} else {
##rawBeta <- .getEstimate(mat = mat, design = design, coef = coef,
##full = FALSE)
##weights <- NULL
##}
###Output
_____no_output_____
###Markdown
.getEstimate() commandsCurrently doesn't handle a user input matrix for the permutations argument (assumes argument has been left blank and defaults to None).Found issues when trying to find corresponding numpy functions for the R matrix operations used in solving the linear regression. The following is a cheatsheet I came up with after experimenting with simple test matrices to see what commands produced the same output in both languages:R | Python --- | --- A %*% B | np.dot(A,B)t(A) | A.Tcrossprod(A,B) | np.dot(A.T,B)tcrossprod(A,B) | np.dot(A,B.T)
###Code
##.getEstimate <- function(mat, design, coef, B=NULL, permutations=NULL,
##full=FALSE) {
##v <- design[,coef]
##A <- design[,-coef, drop=FALSE]
##qa <- qr(A)
##S <- diag(nrow(A)) - tcrossprod(qr.Q(qa))
##vv <- if (is.null(B)) matrix(v, ncol=1) else{
##if(is.null(permutations)){
##replicate(B, sample(v))
##} else{
##apply(permutations,2,function(i) v[i])
##}
##}
##sv <- S %*% vv
##vsv <- diag(crossprod(vv,sv))
##b <- (mat %*% crossprod(S, vv)) / vsv
##if(!is.matrix(b))
##b <- matrix(b, ncol = 1)
##if (full) {
##sy <- mat %*% S
##df.residual <- ncol(mat) - qa$rank - 1
##if (is.null(B)) {
##sigma <- matrix(sqrt(rowSums((sy - tcrossprod(b, sv))^2) / df.residual), ncol=1)
##} else {
##sigma <- b
##tmp <- sy
##for (j in 1:B) {
##tmp <- tcrossprod(b[,j], sv[,j])
##sigma[,j] <- rowSums((sy-tmp)^2)
##}
##sigma <- sqrt(sigma/df.residual)
##}
##out <- list(coef=b,
##sigma=sigma,
##stdev.unscaled=sqrt(1/vsv),
##df.residual=df.residual)
##if (is.null(B)) out$stdev <- as.numeric(out$stdev)
##} else {
##out <- b
##}
##return(out)
##}
##Default values
##coef = 2 --> no need for this in Python implementation
B=None ##Works with B > 0 when tested
permutations=None
full=False ##Works with full = True when tested
##Select the columns of the data frame that are actual phenotype descriptors
v = phenoDesign.loc[:,phenoDesign.columns!="(Intercept)"]
##Select the first, non-phenotype column of the design data frame
Adf = phenoDesign["(Intercept)"]
##Save this one dataframe column as a numpy array
A = np.array(Adf)
##Make sure the array is 2 dimensional
A = A.reshape(-1,1)
##Calculate the QR decomposition of the matrix
####May want to replace with scipy.linalg.qr to get rank more legitimately later
qa = np.linalg.qr(A)
##Create a diagonal matrix with the same dimensions as the number of samples
diagonal = np.zeros((A.shape[0], A.shape[0]), int)
np.fill_diagonal(diagonal, 1)
##Subtract the cross product of qa times the transposition of itself from the diagonal matrix
qa_mat = np.matrix(qa[0])
S = diagonal - np.dot(qa_mat,qa_mat.T)
##If the user didn't specify a number of permutations in the function input
if B is None:
##use the array version of v (the phenotype column(s)) for vv
vv = np.array(v)
else:
##If the user didn't provide a matrix for the permutations argument
if permutations is None:
##convert pandas series v into an array and flatten it to one dimensional
v_arr = np.array(v)
v_arr = v_arr.reshape(v_arr.shape[0])
##Randomly permute the phenotypes among the samples B times
vv = np.random.choice(v_arr,v_arr.shape[0],replace=False)
for i in range(B-1):
next_sample = np.random.choice(v_arr,v_arr.shape[0],replace=False)
vv = np.vstack((vv,next_sample))
vv = vv.T ##Phenotype permutation matrix
vv_mat = np.matrix(vv)
S_mat = np.matrix(S)
sv = np.dot(S_mat,vv_mat)
vsv = np.diag(np.dot(vv_mat.T,sv))
##Bring the matrix of methylation values for each probe into the math
beta_mat = np.matrix(betaVals)
b = np.dot(beta_mat,np.dot(S_mat.T, vv_mat)) / vsv
##If b doesn't turn out to be a matrix, shape the values into a one-column matrix
if type(b) is not np.matrix:
b = np.matrix(b)
b = b.reshape(b.shape[0]*b.shape[1],1)
if full:
sy = np.dot(beta_mat,S_mat)
##Calculate degrees of freedom?
df_residual = beta_mat.shape[1] - qa_mat.shape[1] -1
##If no number of permutations was specified by the user
if B is None:
sigma = np.matrix(np.sqrt(np.sum(np.square(sy-np.dot(b,sv.T)),axis=1)/df_residual))
sigma = sigma.reshape(sigma.shape[0]*sigma.shape[1],1)
else:
sigma = b
tmp = sy
for j in range(B):
tmp = np.dot(b[:,j],sv[:,j].T)
sigma[:,j] = np.sum(np.square(sy-tmp),axis=1)
sigma = np.sqrt(sigma/df_residual)
coef = b
stdev_unscaled = np.sqrt(1/vsv)
##Below are the data structures returned by .getEstimate()
##out <- list(coef=b,
##sigma=sigma,
##stdev.unscaled=sqrt(1/vsv),
##df.residual=df.residual)
##if (is.null(B)) out$stdev <- as.numeric(out$stdev)
##} else {
##out <- b
##}
##return(out)
##}
###Output
_____no_output_____
###Markdown
Proceeding without weights but with smoothingThe output of the previous .getEstimate commands are used as the input rawBeta for the smoothing function.
###Code
##if (smooth) {
##if (verbose)
##message("[bumphunterEngine] Smoothing coefficients.")
##beta <- smoother(y = rawBeta, x = pos, cluster = cluster,
##weights = weights, smoothFunction = smoothFunction,
##verbose = subverbose, ...)
##Index <- which(beta$smoothed)
##beta <- beta$fitted
##} else {
##beta <- rawBeta
##Index <- seq(along = beta)
##}
##smoother <- function(y, x=NULL, cluster, weights=NULL,
##smoothFunction, verbose=TRUE, ...) {
## y = rawBeta --> b output from previous commands
## x = pos --> probePositions.Position
## cluster --> clusterIDs
###Output
_____no_output_____
###Markdown
smoother() function commandsThis function can use parallel processing, but it has NOT been implemented yet in this python translation
###Code
##smoother <- function(y, x=NULL, cluster, weights=NULL,
##smoothFunction, verbose=TRUE, ...) {
weights = None
verbose = True
##if(is.null(dim(y)))
##y <- matrix(y, ncol=1) ##need to change this to be more robust
if type(b) is np.matrix: ##If b is not a matrix
y = b
else: ##reshape its values into a 1 column matrix
y = np.matrix(b)
y = y.reshape(y.shape[0]*y.shape[1],1)
##if(!is.null(weights) && is.null(dim(weights)))
##weights <- matrix(weights, ncol=1)
if (weights is not None) and (type(weights) is not np.matrix):
weights = np.matrix(weights)
weights = weights.reshape(weights.shape[0]*weights.shape[1],1)
##if (!getDoParRegistered())
##registerDoSEQ()
##cores <- getDoParWorkers()
cores = 1
##Indexes <- split(seq(along=cluster), cluster)
##baseSize <- length(Indexes) %/% cores
##remain <- length(Indexes) %% cores
##done <- 0L
##IndexesChunks <- vector("list", length = cores)
Indexes = []
ID_num = clusterIDs[0]
thisInd = [1]
for i in range(1,len(clusterIDs)):
if clusterIDs[i] == ID_num:
thisInd.append(i)
ID_num = clusterIDs[i]
else:
Indexes.append(thisInd)
thisInd = [i]
ID_num = clusterIDs[i]
baseSize = int(len(Indexes)/cores) ##number of clusters to assign to each core
remain = len(Indexes) % cores ##leftover clusters if they didn't divide evenly among cores
done = 0 ##counter to keep track of clusters that have already been assigned to a core
IndexesChunks = [0]*cores ##list to hold each cores' list of clusters
##for(ii in 1:cores) {
##if(remain > 0) { ##
##IndexesChunks[[ii]] <- done + 1:(baseSize + 1)
##remain <- remain - 1L
##done <- done + baseSize + 1L
##} else {
##IndexesChunks[[ii]] <- done + 1:baseSize
##done <- done + baseSize
##}
##}
##For each available core, create a list of
for i in range(cores):
if remain > 0:
baseList = range(baseSize+1)
IndexesChunks[i] = [num+done for num in baseList]
remain -= 1
done += baseSize + 1
else:
baseList = range(baseSize)
IndexesChunks[i] = [num+done for num in baseList]
done += baseSize
##Commands for parallelization with R's foreach
##IndexesChunks <- lapply(IndexesChunks, function(idxes) {
##do.call(c, unname(Indexes[idxes]))
##})
##idx <- NULL ## for R CMD check
##ret <- foreach(idx = iter(IndexesChunks), .packages = "bumphunter") %dorng% {
####APPLY SMOOTHING FUNCTION#####
##sm <- smoothFunction(y=y[idx,], x=x[idx], cluster=cluster[idx],
##weights=weights[idx,], verbose=verbose, ...)
##c(sm, list(idx = idx))
##}
##smoothFunction can be one of three options as specified in the smoother() arguments:
##locfit
##loess
##runmed
if smoothFunction == "loess":
sm.nonparametric.lowess(y,x)
elif smoothFunction == "runmed":
pd.rolling_median(y,window)
##window in this case is an integer --> need to update for each cluster somehow
##attributes(ret)[["rng"]] <- NULL
## Paste together results from different workers
##ret <- reduceIt(ret)
## Now fixing order issue
##revOrder <- ret$idx
##names(revOrder) <- seq_along(ret$idx)
##revOrder <- sort(revOrder)
##revOrder <- as.integer(names(revOrder))
##ret$smoother <- ret$smoother[1]
##ret$fitted <- ret$fitted[revOrder,,drop=FALSE]
##ret$smoothed <- ret$smoothed[revOrder]
##ret$idx <- NULL
##return(ret)
##}
baseSize
###Output
_____no_output_____
###Markdown
Python equivalents for R's smoothing function options used in bumphunter[locfit](http://ugrad.stat.ubc.ca/R/library/locfit/html/locfit.html) - returns a fit object fitted to a local regression model (methods are outlined in the book [Local Regression and Likelihood](http://web.ipac.caltech.edu/staff/fmasci/home/astro_refs/LocalRegressionBook_1999.pdf)) Python options: * [Python wrapper for Locfit C routines](https://github.com/chairmank/python-locfit) that uses numpy conventions in a github repository. Hasn't been updated in years, but says it has an MIT Open Source Software license (link is dead).* Github project for a [Gradient Boosting Machine that uses Locfit](https://github.com/materialsproject/gbml/) for local regression. It's been updated recently, but hard to tell how you actually use it in Python... Seems to rely on wrapper project above - also uses local regression to smooth polynomial Python options: * ~~An individual wrote up [code for a LOESS non-parametric smoothing function](http://www.jtrive.com/loess-nonparametric-scatterplot-smoothing-in-python.html) (dependencies: pandas, numpy, scipy)~~* A function for lowess smoothing is available [in the statsmodel module](https://www.statsmodels.org/dev/generated/statsmodels.nonparametric.smoothers_lowess.lowess.html)* ~~The scikit-misc module also offers a [loess function](https://has2k1.github.io/scikit-misc/loess.html).~~* ~~Another individual on github made a [lowess smoothing script](https://gist.github.com/agramfort/850437). (dependencies: numpy, scipy)~~[runmed](https://www.rdocumentation.org/packages/stats/versions/3.5.1/topics/runmed) - calculates a running median across a list of values for a given window Python options:* pandas has a [rolling median function](https://pandas.pydata.org/pandas-docs/version/0.17.0/generated/pandas.rolling_median.html), sold! smoothFunction() commands --> loessByCluster() smoothingIn the middle of smoother, the R function calls one of three different smootherFunctions depending on user specification. The default is locfit, but another option is loess which is also based on local regression.
###Code
##sm <- smoothFunction(y=y[idx,], x=x[idx], cluster=cluster[idx],
##weights=weights[idx,], verbose=verbose, ...)
##loessByCluster <- function(y, x=NULL, cluster, weights= NULL,
##bpSpan = 1000, minNum=7, minInSpan=5,
##maxSpan=1, verbose=TRUE)
# x = chromosome positions
# y = regression beta values
## Depends on limma
## bpSpan is in basepairs
## assumed x are ordered
## if y is vector change to matrix
x = probePositions.Position
y = b
bpSpan = 1000
minNum = 7 ##minimum number of probes in a cluster to be able to smooth the cluster
minInSpan = 5
maxSpan = 1
verbose = True
##if(is.null(dim(y)))
##y <- matrix(y, ncol=1) ##need to change this to be more robust
##if(!is.null(weights) && is.null(dim(weights)))
##weights <- matrix(weights,ncol=1)
##if(is.null(x))
##x <- seq(along=y)
if x is None: ##if no x variable is provided, use a sequential list the length of y
x = range(y.shape[0])
##if(is.null(weights))
##weights <- matrix(1, nrow=nrow(y), ncol=ncol(y))
if weights is None: ##if no weights are provided, make a matrix the same size as y
weights = np.zeros((y.shape[0],y.shape[1]))
weights += 1 ##with every weight equal to 1
##Indexes <- split(seq(along=cluster), cluster) --> done in cell above for smoother()
##clusterL <- sapply(Indexes, length)
##spans <- rep(NA, length(Indexes))
##smoothed <- rep(TRUE, nrow(y))
clusterL = [] ##list that counts the number of probes in each cluster
for i in range(len(Indexes)):
clusterL.append(len(Indexes[i]))
spans = [None]*len(Indexes)
smoothed = [True]*y.shape[0]
##for(i in seq(along=Indexes)) {
for i in range(len(Indexes)):
##if(verbose) if(i %% 1e4 == 0) cat(".")
##Index <- Indexes[[i]]
Index = Indexes[i]
##if(clusterL[i] >= minNum) {
##if this cluster contains the minimum number of probes required for smoothing
if (clusterL[i] >= minNum)
##span = bpSpan/median(diff(x[Index]))/length(Index) # make a span
##if(span > maxSpan) span <- maxSpan
##spans[i] <- span
span = bpSpan/np.median(np.diff(x[Index]))/len(Index)
if span > maxSpan:
span = maxSpan
spans[i] = span
##if(span*length(Index)>minInSpan){
##this can be parallelized
##for(j in 1:ncol(y)){
##y[Index,j] <- limma::loessFit(y[Index,j], x[Index], span = span,
##weights = weights[Index,j])$fitted
##}
if span*len(Index) > minInSpan:
for j in range(y.shape[0]):
y[Index,j]
##} else{
##y[Index,] <- NA
##smoothed[Index] <- FALSE
##}
##} else{
##y[Index,] <- NA
##spans[i] <- NA
##smoothed[Index] <- FALSE
##}
##}
##return(list(fitted=y, smoothed=smoothed, smoother="loess"))
##}
x = probePositions.Position
##span = bpSpan/median(diff(x[Index]))/length(Index) # make a span
##if(span > maxSpan) span <- maxSpan
##spans[i] <- span
Index = Indexes[16]
bpSpan=1000
bpSpan/np.median(np.diff(x[Index]))/len(Index)
coef.shape
###Output
_____no_output_____
###Markdown
smoothFunction() commands --> runmed smoothing
###Code
clusterIDs[:50]
weights
###Output
_____no_output_____ |
P1/actors_and_movies2.0.ipynb | ###Markdown
We have duplicated movies in some actors
###Code
for i in range(df["Movies"].count()):
df["Movies"][i] = list(set(df["Movies"][i])) #Can't use pandas.unique() cos type of ["Movies"][i] is -> list
df["Movies"][18]
###Output
_____no_output_____
###Markdown
We already eliminate duplicates and get the df to a correct format so we cant start with: Graph analysis with networkx
###Code
import networkx as nx
G = nx.Graph()
start_time = time.time() #Compute time of execution
for i in range(df["Movies"].count()):
G.add_nodes_from(df["Movies"][i]) #1896 nodes
def add_all_edges(G, total_actors):
for i in range(total_actors):
for j in range(len(df["Movies"][i])-1):
#Dont go over the full list -> quicker
#why? -> last element connected in previous iter
for k in range(len(df["Movies"][i])):
if(j >= k): #Dont add equal edges or already added
str1 = df["Movies"][i][j]
str2 = df["Movies"][i][k]
G.add_edge(str1,str2)
start_time = time.time() #Compute time of execution
total_actors = df["Movies"].count()
#total_actors = int(total_actors*0.01) #Gets 20% of the actors, comment out to get full set
print("Nº Actors:", total_actors)
add_all_edges(G, total_actors)
print("--- %s seconds ---" % (time.time() - start_time))
print("Nº Nodes:", G.number_of_nodes())
print("Nº Edges:",G.number_of_edges())
degrees = nx.degree(G)
degrees = dict(degrees)
mean_deg = []
for k,v in degrees.items():
mean_deg.append(v)
if(v == 0):
G.remove_node(k)
print("\n--- After degree 0 removing ---")
print("Nº Nodes:", G.number_of_nodes())
print("Nº Edges:",G.number_of_edges())
G_less = G.copy()
G_great = G.copy()
for k,v in degrees.items():
if( v != 0):
if(v > 5):
G_less.remove_node(k)
if(v < 30):
G_great.remove_node(k)
#Quitar nodos con menos de x enlaces
print(G_less.number_of_nodes())
print(G_great.number_of_nodes())
#Quitar nodos con muchos enlaces
###Output
389
244
|
machine_learning/gan/cgan/tf_cgan/tf_cgan_run_module_local.ipynb | ###Markdown
Run model module locally
###Code
import os
# Import os environment variables for file hyperparameters.
os.environ["TRAIN_FILE_PATTERN"] = "gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord"
os.environ["EVAL_FILE_PATTERN"] = "gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord"
os.environ["OUTPUT_DIR"] = "gs://machine-learning-1234-bucket/gan/cgan/trained_model2"
# Import os environment variables for train hyperparameters.
os.environ["TRAIN_BATCH_SIZE"] = str(100)
os.environ["TRAIN_STEPS"] = str(60000)
os.environ["SAVE_SUMMARY_STEPS"] = str(100)
os.environ["SAVE_CHECKPOINTS_STEPS"] = str(5000)
os.environ["KEEP_CHECKPOINT_MAX"] = str(10)
os.environ["INPUT_FN_AUTOTUNE"] = "False"
# Import os environment variables for eval hyperparameters.
os.environ["EVAL_BATCH_SIZE"] = str(16)
os.environ["EVAL_STEPS"] = str(10)
os.environ["START_DELAY_SECS"] = str(6000)
os.environ["THROTTLE_SECS"] = str(6000)
# Import os environment variables for image hyperparameters.
os.environ["HEIGHT"] = str(28)
os.environ["WIDTH"] = str(28)
os.environ["DEPTH"] = str(1)
# Import os environment variables for label hyperparameters.
num_classes = 10
os.environ["NUM_CLASSES"] = str(num_classes)
os.environ["LABEL_EMBEDDING_DIMENSION"] = str(10)
# Import os environment variables for generator hyperparameters.
os.environ["LATENT_SIZE"] = str(512)
os.environ["GENERATOR_USE_LABELS"] = "True"
os.environ["GENERATOR_EMBED_LABELS"] = "True"
os.environ["GENERATOR_CONCATENATE_LABELS"] = "True"
os.environ["GENERATOR_DENSE_BEFORE_CONCATENATE"] = "False"
os.environ["GENERATOR_HIDDEN_UNITS"] = "256,512,1024"
os.environ["GENERATOR_LEAKY_RELU_ALPHA"] = str(0.2)
os.environ["GENERATOR_FINAL_ACTIVATION"] = "tanh"
os.environ["GENERATOR_L1_REGULARIZATION_SCALE"] = str(0.)
os.environ["GENERATOR_L2_REGULARIZATION_SCALE"] = str(0.)
os.environ["GENERATOR_OPTIMIZER"] = "Adam"
os.environ["GENERATOR_LEARNING_RATE"] = str(0.0002)
os.environ["GENERATOR_ADAM_BETA1"] = str(0.5)
os.environ["GENERATOR_ADAM_BETA2"] = str(0.999)
os.environ["GENERATOR_ADAM_EPSILON"] = str(1e-8)
os.environ["GENERATOR_CLIP_GRADIENTS"] = "None"
os.environ["GENERATOR_TRAIN_STEPS"] = str(1)
# Import os environment variables for discriminator hyperparameters.
os.environ["DISCRIMINATOR_USE_LABELS"] = "True"
os.environ["DISCRIMINATOR_EMBED_LABELS"] = "True"
os.environ["DISCRIMINATOR_CONCATENATE_LABELS"] = "True"
os.environ["DISCRIMINATOR_DENSE_BEFORE_CONCATENATE"] = "False"
os.environ["DISCRIMINATOR_HIDDEN_UNITS"] = "1024,512,256"
os.environ["DISCRIMINATOR_LEAKY_RELU_ALPHA"] = str(0.2)
os.environ["DISCRIMINATOR_L1_REGULARIZATION_SCALE"] = str(0.)
os.environ["DISCRIMINATOR_L2_REGULARIZATION_SCALE"] = str(0.)
os.environ["DISCRIMINATOR_OPTIMIZER"] = "Adam"
os.environ["DISCRIMINATOR_LEARNING_RATE"] = str(0.0002)
os.environ["DISCRIMINATOR_ADAM_BETA1"] = str(0.5)
os.environ["DISCRIMINATOR_ADAM_BETA2"] = str(0.999)
os.environ["DISCRIMINATOR_ADAM_EPSILON"] = str(1e-8)
os.environ["DISCRIMINATOR_CLIP_GRADIENTS"] = "None"
os.environ["DISCRIMINATOR_TRAIN_STEPS"] = str(1)
os.environ["LABEL_SMOOTHING"] = str(0.9)
###Output
_____no_output_____
###Markdown
Train Vanilla GAN model
###Code
%%bash
gsutil -m rm -rf ${OUTPUT_DIR}
export PYTHONPATH=$PYTHONPATH:$PWD/cgan_module
python3 -m trainer.task \
--train_file_pattern=${TRAIN_FILE_PATTERN} \
--eval_file_pattern=${EVAL_FILE_PATTERN} \
--output_dir=${OUTPUT_DIR} \
--job-dir=./tmp \
\
--train_batch_size=${TRAIN_BATCH_SIZE} \
--train_steps=${TRAIN_STEPS} \
--save_summary_steps=${SAVE_SUMMARY_STEPS} \
--save_checkpoints_steps=${SAVE_CHECKPOINTS_STEPS} \
--keep_checkpoint_max=${KEEP_CHECKPOINT_MAX} \
--input_fn_autotune=${INPUT_FN_AUTOTUNE} \
\
--eval_batch_size=${EVAL_BATCH_SIZE} \
--eval_steps=${EVAL_STEPS} \
--start_delay_secs=${START_DELAY_SECS} \
--throttle_secs=${THROTTLE_SECS} \
\
--height=${HEIGHT} \
--width=${WIDTH} \
--depth=${DEPTH} \
\
--num_classes=${NUM_CLASSES} \
--label_embedding_dimension=${LABEL_EMBEDDING_DIMENSION} \
\
--latent_size=${LATENT_SIZE} \
--generator_use_labels=${GENERATOR_USE_LABELS} \
--generator_embed_labels=${GENERATOR_EMBED_LABELS} \
--generator_concatenate_labels=${GENERATOR_CONCATENATE_LABELS} \
--generator_dense_before_concatenate=${GENERATOR_DENSE_BEFORE_CONCATENATE} \
--generator_hidden_units=${GENERATOR_HIDDEN_UNITS} \
--generator_leaky_relu_alpha=${GENERATOR_LEAKY_RELU_ALPHA} \
--generator_final_activation=${GENERATOR_FINAL_ACTIVATION} \
--generator_l1_regularization_scale=${GENERATOR_L1_REGULARIZATION_SCALE} \
--generator_l2_regularization_scale=${GENERATOR_L2_REGULARIZATION_SCALE} \
--generator_optimizer=${GENERATOR_OPTIMIZER} \
--generator_learning_rate=${GENERATOR_LEARNING_RATE} \
--generator_adam_beta1=${GENERATOR_ADAM_BETA1} \
--generator_adam_beta2=${GENERATOR_ADAM_BETA2} \
--generator_adam_epsilon=${GENERATOR_ADAM_EPSILON} \
--generator_clip_gradients=${GENERATOR_CLIP_GRADIENTS} \
--generator_train_steps=${GENERATOR_TRAIN_STEPS} \
\
--discriminator_use_labels=${DISCRIMINATOR_USE_LABELS} \
--discriminator_embed_labels=${DISCRIMINATOR_EMBED_LABELS} \
--discriminator_concatenate_labels=${DISCRIMINATOR_CONCATENATE_LABELS} \
--discriminator_dense_before_concatenate=${DISCRIMINATOR_DENSE_BEFORE_CONCATENATE} \
--discriminator_hidden_units=${DISCRIMINATOR_HIDDEN_UNITS} \
--discriminator_leaky_relu_alpha=${DISCRIMINATOR_LEAKY_RELU_ALPHA} \
--discriminator_l1_regularization_scale=${DISCRIMINATOR_L1_REGULARIZATION_SCALE} \
--discriminator_l2_regularization_scale=${DISCRIMINATOR_L2_REGULARIZATION_SCALE} \
--discriminator_optimizer=${DISCRIMINATOR_OPTIMIZER} \
--discriminator_learning_rate=${DISCRIMINATOR_LEARNING_RATE} \
--discriminator_adam_beta1=${DISCRIMINATOR_ADAM_BETA1} \
--discriminator_adam_beta2=${DISCRIMINATOR_ADAM_BETA2} \
--discriminator_adam_epsilon=${DISCRIMINATOR_ADAM_EPSILON} \
--discriminator_clip_gradients=${DISCRIMINATOR_CLIP_GRADIENTS} \
--discriminator_train_steps=${DISCRIMINATOR_TRAIN_STEPS} \
--label_smoothing=${LABEL_SMOOTHING}
###Output
train_and_evaluate: args = {'train_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord', 'eval_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord', 'output_dir': 'gs://machine-learning-1234-bucket/gan/cgan/trained_model2/', 'train_batch_size': 100, 'train_steps': 60000, 'save_summary_steps': 100, 'save_checkpoints_steps': 5000, 'keep_checkpoint_max': 10, 'eval_batch_size': 16, 'eval_steps': 10, 'start_delay_secs': 6000, 'throttle_secs': 6000, 'height': 28, 'width': 28, 'depth': 1, 'num_classes': 10, 'label_embedding_dimension': 10, 'latent_size': 512, 'generator_use_labels': True, 'generator_embed_labels': True, 'generator_concatenate_labels': True, 'generator_dense_before_concatenate': False, 'generator_hidden_units': [256, 512, 1024], 'generator_leaky_relu_alpha': 0.2, 'generator_final_activation': 'tanh', 'generator_l1_regularization_scale': 0.0, 'generator_l2_regularization_scale': 0.0, 'generator_optimizer': 'Adam', 'generator_learning_rate': 0.0002, 'generator_adam_beta1': 0.5, 'generator_adam_beta2': 0.999, 'generator_adam_epsilon': 1e-08, 'generator_clip_gradients': None, 'generator_train_steps': 1, 'discriminator_use_labels': True, 'discriminator_embed_labels': True, 'discriminator_concatenate_labels': True, 'discriminator_dense_before_concatenate': False, 'discriminator_hidden_units': [1024, 512, 256], 'discriminator_leaky_relu_alpha': 0.2, 'discriminator_l1_regularization_scale': 0.0, 'discriminator_l2_regularization_scale': 0.0, 'discriminator_optimizer': 'Adam', 'discriminator_learning_rate': 0.0002, 'discriminator_adam_beta1': 0.5, 'discriminator_adam_beta2': 0.999, 'discriminator_adam_epsilon': 1e-08, 'discriminator_clip_gradients': None, 'discriminator_train_steps': 1, 'label_smoothing': 0.9}
decode_example: features = {'image_raw': FixedLenFeature(shape=[], dtype=tf.string, default_value=None), 'label': FixedLenFeature(shape=[], dtype=tf.int64, default_value=None)}
decode_example: image = Tensor("DecodeRaw:0", shape=(?,), dtype=uint8)
decode_example: image = Tensor("Reshape:0", shape=(28, 28, 1), dtype=uint8)
preprocess_image: image = Tensor("sub:0", shape=(28, 28, 1), dtype=float32)
decode_example: image = Tensor("sub:0", shape=(28, 28, 1), dtype=float32)
decode_example: label = Tensor("Cast_1:0", shape=(), dtype=int32)
cgan_model: features = {'image': <tf.Tensor 'IteratorGetNext:0' shape=(?, 28, 28, 1) dtype=float32>}
cgan_model: labels = Tensor("IteratorGetNext:1", shape=(?,), dtype=int32, device=/device:CPU:0)
cgan_model: mode = train
cgan_model: params = {'train_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord', 'eval_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord', 'output_dir': 'gs://machine-learning-1234-bucket/gan/cgan/trained_model2/', 'train_batch_size': 100, 'train_steps': 60000, 'save_summary_steps': 100, 'save_checkpoints_steps': 5000, 'keep_checkpoint_max': 10, 'eval_batch_size': 16, 'eval_steps': 10, 'start_delay_secs': 6000, 'throttle_secs': 6000, 'height': 28, 'width': 28, 'depth': 1, 'num_classes': 10, 'label_embedding_dimension': 10, 'latent_size': 512, 'generator_use_labels': True, 'generator_embed_labels': True, 'generator_concatenate_labels': True, 'generator_dense_before_concatenate': False, 'generator_hidden_units': [256, 512, 1024], 'generator_leaky_relu_alpha': 0.2, 'generator_final_activation': 'tanh', 'generator_l1_regularization_scale': 0.0, 'generator_l2_regularization_scale': 0.0, 'generator_optimizer': 'Adam', 'generator_learning_rate': 0.0002, 'generator_adam_beta1': 0.5, 'generator_adam_beta2': 0.999, 'generator_adam_epsilon': 1e-08, 'generator_clip_gradients': None, 'generator_train_steps': 1, 'discriminator_use_labels': True, 'discriminator_embed_labels': True, 'discriminator_concatenate_labels': True, 'discriminator_dense_before_concatenate': False, 'discriminator_hidden_units': [1024, 512, 256], 'discriminator_leaky_relu_alpha': 0.2, 'discriminator_l1_regularization_scale': 0.0, 'discriminator_l2_regularization_scale': 0.0, 'discriminator_optimizer': 'Adam', 'discriminator_learning_rate': 0.0002, 'discriminator_adam_beta1': 0.5, 'discriminator_adam_beta2': 0.999, 'discriminator_adam_epsilon': 1e-08, 'discriminator_clip_gradients': None, 'discriminator_train_steps': 1, 'label_smoothing': 0.9}
cgan_model: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
Training discriminator.
get_logits_and_losses: real_images = Tensor("real_images:0", shape=(?, 784), dtype=float32)
get_logits_and_losses: Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32)
Call generator with Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32).
get_fake_images: Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
Call discriminator with fake_images = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
Call discriminator with real_images = Tensor("real_images:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("real_images:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator_1/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator_1/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator_1/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator_1/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
get_discriminator_loss: discriminator_real_loss = Tensor("discriminator_real_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_fake_loss = Tensor("discriminator_fake_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_loss = Tensor("discriminator_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_reg_loss = Tensor("Const_3:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_total_loss = Tensor("discriminator_total_loss:0", shape=(), dtype=float32)
Training generator.
get_logits_and_losses: fake_labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
Call generator with fake_Z = Tensor("generator_Z:0", shape=(?, 512), dtype=float32).
get_fake_images: Z = Tensor("generator_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator_1/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator_1/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator_1/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
Call discriminator with fake_fake_images = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator_2/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator_2/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator_2/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator_2/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
get_generator_loss: generator_loss = Tensor("generator_loss:0", shape=(), dtype=float32)
get_generator_loss: generator_reg_loss = Tensor("Const_5:0", shape=(), dtype=float32)
get_generator_loss: generator_total_loss = Tensor("generator_total_loss:0", shape=(), dtype=float32)
get_fake_images: Z = Tensor("image_summary_Z:0", shape=(10, 512), dtype=float32)
get_fake_images: labels = Tensor("image_summary_fake_labels:0", shape=(10, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator_2/generator/generator/label_vectors:0", shape=(10, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator_2/generator/generator/label_vectors:0", shape=(10, 10), dtype=float32)
generator_use_labels: network = Tensor("generator_2/generator/concat_labels:0", shape=(10, 522), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_0/BiasAdd:0", shape=(10, 256), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_0:0", shape=(10, 256), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_1/BiasAdd:0", shape=(10, 512), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_1:0", shape=(10, 512), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_2/BiasAdd:0", shape=(10, 1024), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_2:0", shape=(10, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator_2/layers_dense_generated_outputs/Tanh:0", shape=(10, 784), dtype=float32)
get_variables_and_gradients_generator: variables = [<tf.Variable 'generator/generator/generator/label_embedding_matrix:0' shape=(10, 10) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_0/kernel:0' shape=(522, 256) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_0/bias:0' shape=(256,) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_1/kernel:0' shape=(256, 512) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_1/bias:0' shape=(512,) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_2/kernel:0' shape=(512, 1024) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_2/bias:0' shape=(1024,) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_generated_outputs/kernel:0' shape=(1024, 784) dtype=float32_ref>, <tf.Variable 'generator/layers_dense_generated_outputs/bias:0' shape=(784,) dtype=float32_ref>]
get_variables_and_gradients_generator: gradients = [<tensorflow.python.framework.indexed_slices.IndexedSlices object at 0x7f896676df90>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_0/MatMul_grad/MatMul_1:0' shape=(522, 256) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_0/BiasAdd_grad/BiasAddGrad:0' shape=(256,) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_1/MatMul_grad/MatMul_1:0' shape=(256, 512) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_1/BiasAdd_grad/BiasAddGrad:0' shape=(512,) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_2/MatMul_grad/MatMul_1:0' shape=(512, 1024) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_2/BiasAdd_grad/BiasAddGrad:0' shape=(1024,) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_generated_outputs/MatMul_grad/MatMul_1:0' shape=(1024, 784) dtype=float32>, <tf.Tensor 'generator_gradients/generator_1/layers_dense_generated_outputs/BiasAdd_grad/BiasAddGrad:0' shape=(784,) dtype=float32>]
get_variables_and_gradients_generator: gradients = [<tf.Tensor 'get_variables_and_gradients_generator/generator/generator/label_embedding_matrix_gradients:0' shape=(10, 10) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_0/kernel_gradients:0' shape=(522, 256) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_0/bias_gradients:0' shape=(256,) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_1/kernel_gradients:0' shape=(256, 512) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_1/bias_gradients:0' shape=(512,) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_2/kernel_gradients:0' shape=(512, 1024) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_2/bias_gradients:0' shape=(1024,) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_generated_outputs/kernel_gradients:0' shape=(1024, 784) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_generator/layers_dense_generated_outputs/bias_gradients:0' shape=(784,) dtype=float32>]
get_variables_and_gradients_discriminator: variables = [<tf.Variable 'discriminator/discriminator/discriminator/label_embedding_matrix:0' shape=(10, 10) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_0/kernel:0' shape=(794, 1024) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_0/bias:0' shape=(1024,) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_1/kernel:0' shape=(1024, 512) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_1/bias:0' shape=(512,) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_2/kernel:0' shape=(512, 256) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_2/bias:0' shape=(256,) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_logits/kernel:0' shape=(256, 1) dtype=float32_ref>, <tf.Variable 'discriminator/layers_dense_logits/bias:0' shape=(1,) dtype=float32_ref>]
get_variables_and_gradients_discriminator: gradients = [<tensorflow.python.framework.indexed_slices.IndexedSlices object at 0x7f89666fbd10>, <tf.Tensor 'discriminator_gradients/AddN_9:0' shape=(794, 1024) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_8:0' shape=(1024,) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_7:0' shape=(1024, 512) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_6:0' shape=(512,) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_5:0' shape=(512, 256) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_4:0' shape=(256,) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_3:0' shape=(256, 1) dtype=float32>, <tf.Tensor 'discriminator_gradients/AddN_2:0' shape=(1,) dtype=float32>]
get_variables_and_gradients_discriminator: gradients = [<tf.Tensor 'get_variables_and_gradients_discriminator/discriminator/discriminator/label_embedding_matrix_gradients:0' shape=(10, 10) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_0/kernel_gradients:0' shape=(794, 1024) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_0/bias_gradients:0' shape=(1024,) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_1/kernel_gradients:0' shape=(1024, 512) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_1/bias_gradients:0' shape=(512,) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_2/kernel_gradients:0' shape=(512, 256) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_2/bias_gradients:0' shape=(256,) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_logits/kernel_gradients:0' shape=(256, 1) dtype=float32>, <tf.Tensor 'get_variables_and_gradients_discriminator/layers_dense_logits/bias_gradients:0' shape=(1,) dtype=float32>]
train_network: scope = discriminator
train_network_discriminator: optimizer = <tensorflow.python.training.adam.AdamOptimizer object at 0x7f89666ac3d0>
train_network_discriminator: gradients = [<tensorflow.python.framework.indexed_slices.IndexedSlices object at 0x7f89666c3d90>, <tf.Tensor 'cond/discriminator_gradients/AddN_9:0' shape=(794, 1024) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_8:0' shape=(1024,) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_7:0' shape=(1024, 512) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_6:0' shape=(512,) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_5:0' shape=(512, 256) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_4:0' shape=(256,) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_3:0' shape=(256, 1) dtype=float32>, <tf.Tensor 'cond/discriminator_gradients/AddN_2:0' shape=(1,) dtype=float32>]
train_network_discriminator: grads_and_vars = <zip object at 0x7f89666ba320>
train_network: scope = generator
train_network_generator: optimizer = <tensorflow.python.training.adam.AdamOptimizer object at 0x7f89666ac3d0>
train_network_generator: gradients = [<tensorflow.python.framework.indexed_slices.IndexedSlices object at 0x7f8966501250>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_0/MatMul_grad/MatMul_1:0' shape=(522, 256) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_0/BiasAdd_grad/BiasAddGrad:0' shape=(256,) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_1/MatMul_grad/MatMul_1:0' shape=(256, 512) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_1/BiasAdd_grad/BiasAddGrad:0' shape=(512,) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_2/MatMul_grad/MatMul_1:0' shape=(512, 1024) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_2/BiasAdd_grad/BiasAddGrad:0' shape=(1024,) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_generated_outputs/MatMul_grad/MatMul_1:0' shape=(1024, 784) dtype=float32>, <tf.Tensor 'cond/generator_gradients/generator_1/layers_dense_generated_outputs/BiasAdd_grad/BiasAddGrad:0' shape=(784,) dtype=float32>]
train_network_generator: grads_and_vars = <zip object at 0x7f896662d5a0>
decode_example: features = {'image_raw': FixedLenFeature(shape=[], dtype=tf.string, default_value=None), 'label': FixedLenFeature(shape=[], dtype=tf.int64, default_value=None)}
decode_example: image = Tensor("DecodeRaw:0", shape=(?,), dtype=uint8)
decode_example: image = Tensor("Reshape:0", shape=(28, 28, 1), dtype=uint8)
preprocess_image: image = Tensor("sub:0", shape=(28, 28, 1), dtype=float32)
decode_example: image = Tensor("sub:0", shape=(28, 28, 1), dtype=float32)
decode_example: label = Tensor("Cast_1:0", shape=(), dtype=int32)
cgan_model: features = {'image': <tf.Tensor 'IteratorGetNext:0' shape=(?, 28, 28, 1) dtype=float32>}
cgan_model: labels = Tensor("IteratorGetNext:1", shape=(?,), dtype=int32, device=/device:CPU:0)
cgan_model: mode = eval
cgan_model: params = {'train_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord', 'eval_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord', 'output_dir': 'gs://machine-learning-1234-bucket/gan/cgan/trained_model2/', 'train_batch_size': 100, 'train_steps': 60000, 'save_summary_steps': 100, 'save_checkpoints_steps': 5000, 'keep_checkpoint_max': 10, 'eval_batch_size': 16, 'eval_steps': 10, 'start_delay_secs': 6000, 'throttle_secs': 6000, 'height': 28, 'width': 28, 'depth': 1, 'num_classes': 10, 'label_embedding_dimension': 10, 'latent_size': 512, 'generator_use_labels': True, 'generator_embed_labels': True, 'generator_concatenate_labels': True, 'generator_dense_before_concatenate': False, 'generator_hidden_units': [256, 512, 1024], 'generator_leaky_relu_alpha': 0.2, 'generator_final_activation': 'tanh', 'generator_l1_regularization_scale': 0.0, 'generator_l2_regularization_scale': 0.0, 'generator_optimizer': 'Adam', 'generator_learning_rate': 0.0002, 'generator_adam_beta1': 0.5, 'generator_adam_beta2': 0.999, 'generator_adam_epsilon': 1e-08, 'generator_clip_gradients': None, 'generator_train_steps': 1, 'discriminator_use_labels': True, 'discriminator_embed_labels': True, 'discriminator_concatenate_labels': True, 'discriminator_dense_before_concatenate': False, 'discriminator_hidden_units': [1024, 512, 256], 'discriminator_leaky_relu_alpha': 0.2, 'discriminator_l1_regularization_scale': 0.0, 'discriminator_l2_regularization_scale': 0.0, 'discriminator_optimizer': 'Adam', 'discriminator_learning_rate': 0.0002, 'discriminator_adam_beta1': 0.5, 'discriminator_adam_beta2': 0.999, 'discriminator_adam_epsilon': 1e-08, 'discriminator_clip_gradients': None, 'discriminator_train_steps': 1, 'label_smoothing': 0.9}
cgan_model: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
Training discriminator.
get_logits_and_losses: real_images = Tensor("real_images:0", shape=(?, 784), dtype=float32)
get_logits_and_losses: Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32)
Call generator with Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32).
get_fake_images: Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
Call discriminator with fake_images = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
Call discriminator with real_images = Tensor("real_images:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("real_images:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator_1/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator_1/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator_1/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator_1/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
get_discriminator_loss: discriminator_real_loss = Tensor("discriminator_real_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_fake_loss = Tensor("discriminator_fake_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_loss = Tensor("discriminator_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_reg_loss = Tensor("Const_3:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_total_loss = Tensor("discriminator_total_loss:0", shape=(), dtype=float32)
Training generator.
get_logits_and_losses: fake_labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
Call generator with fake_Z = Tensor("generator_Z:0", shape=(?, 512), dtype=float32).
get_fake_images: Z = Tensor("generator_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator_1/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator_1/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator_1/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
Call discriminator with fake_fake_images = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator_2/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator_2/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator_2/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator_2/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
get_generator_loss: generator_loss = Tensor("generator_loss:0", shape=(), dtype=float32)
get_generator_loss: generator_reg_loss = Tensor("Const_5:0", shape=(), dtype=float32)
get_generator_loss: generator_total_loss = Tensor("generator_total_loss:0", shape=(), dtype=float32)
get_fake_images: Z = Tensor("image_summary_Z:0", shape=(10, 512), dtype=float32)
get_fake_images: labels = Tensor("image_summary_fake_labels:0", shape=(10, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator_2/generator/generator/label_vectors:0", shape=(10, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator_2/generator/generator/label_vectors:0", shape=(10, 10), dtype=float32)
generator_use_labels: network = Tensor("generator_2/generator/concat_labels:0", shape=(10, 522), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_0/BiasAdd:0", shape=(10, 256), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_0:0", shape=(10, 256), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_1/BiasAdd:0", shape=(10, 512), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_1:0", shape=(10, 512), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_2/BiasAdd:0", shape=(10, 1024), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_2:0", shape=(10, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator_2/layers_dense_generated_outputs/Tanh:0", shape=(10, 784), dtype=float32)
get_eval_metric_ops: discriminator_logits = Tensor("discriminator_concat_logits:0", shape=(?, 1), dtype=float32)
get_eval_metric_ops: discriminator_labels = Tensor("discriminator_concat_labels:0", shape=(?, 1), dtype=float32)
get_eval_metric_ops: discriminator_probabilities = Tensor("discriminator_probabilities:0", shape=(?, 1), dtype=float32)
get_eval_metric_ops: eval_metric_ops = {'accuracy': (<tf.Tensor 'discriminator_accuracy/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_accuracy/update_op:0' shape=() dtype=float32>), 'precision': (<tf.Tensor 'discriminator_precision/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_precision/update_op:0' shape=() dtype=float32>), 'recall': (<tf.Tensor 'discriminator_recall/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_recall/update_op:0' shape=() dtype=float32>), 'auc_roc': (<tf.Tensor 'discriminator_auc_roc/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_auc_roc/update_op:0' shape=() dtype=float32>), 'auc_pr': (<tf.Tensor 'discriminator_auc_pr/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_auc_pr/update_op:0' shape=() dtype=float32>)}
serving_input_fn: feature_placeholders = {'Z': <tf.Tensor 'serving_input_placeholder_Z:0' shape=(?, 512) dtype=float32>, 'label': <tf.Tensor 'serving_input_placeholder_label:0' shape=(?,) dtype=int32>}
serving_input_fn: features = {'Z': <tf.Tensor 'serving_input_fn_identity_placeholder_Z:0' shape=(?, 512) dtype=float32>, 'label': <tf.Tensor 'serving_input_fn_identity_placeholder_label:0' shape=(?,) dtype=int32>}
cgan_model: features = {'Z': <tf.Tensor 'serving_input_fn_identity_placeholder_Z:0' shape=(?, 512) dtype=float32>, 'label': <tf.Tensor 'serving_input_fn_identity_placeholder_label:0' shape=(?,) dtype=int32>}
cgan_model: labels = None
cgan_model: mode = infer
cgan_model: params = {'train_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord', 'eval_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord', 'output_dir': 'gs://machine-learning-1234-bucket/gan/cgan/trained_model2/', 'train_batch_size': 100, 'train_steps': 60000, 'save_summary_steps': 100, 'save_checkpoints_steps': 5000, 'keep_checkpoint_max': 10, 'eval_batch_size': 16, 'eval_steps': 10, 'start_delay_secs': 6000, 'throttle_secs': 6000, 'height': 28, 'width': 28, 'depth': 1, 'num_classes': 10, 'label_embedding_dimension': 10, 'latent_size': 512, 'generator_use_labels': True, 'generator_embed_labels': True, 'generator_concatenate_labels': True, 'generator_dense_before_concatenate': False, 'generator_hidden_units': [256, 512, 1024], 'generator_leaky_relu_alpha': 0.2, 'generator_final_activation': 'tanh', 'generator_l1_regularization_scale': 0.0, 'generator_l2_regularization_scale': 0.0, 'generator_optimizer': 'Adam', 'generator_learning_rate': 0.0002, 'generator_adam_beta1': 0.5, 'generator_adam_beta2': 0.999, 'generator_adam_epsilon': 1e-08, 'generator_clip_gradients': None, 'generator_train_steps': 1, 'discriminator_use_labels': True, 'discriminator_embed_labels': True, 'discriminator_concatenate_labels': True, 'discriminator_dense_before_concatenate': False, 'discriminator_hidden_units': [1024, 512, 256], 'discriminator_leaky_relu_alpha': 0.2, 'discriminator_l1_regularization_scale': 0.0, 'discriminator_l2_regularization_scale': 0.0, 'discriminator_optimizer': 'Adam', 'discriminator_learning_rate': 0.0002, 'discriminator_adam_beta1': 0.5, 'discriminator_adam_beta2': 0.999, 'discriminator_adam_epsilon': 1e-08, 'discriminator_clip_gradients': None, 'discriminator_train_steps': 1, 'label_smoothing': 0.9}
get_predictions_and_export_outputs: Z = Tensor("serving_input_fn_identity_placeholder_Z:0", shape=(?, 512), dtype=float32)
get_predictions_and_export_outputs: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
get_fake_images: Z = Tensor("serving_input_fn_identity_placeholder_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_predictions_and_export_outputs: fake_images = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_predictions_and_export_outputs: generated_images = Tensor("Reshape:0", shape=(?, 28, 28, 1), dtype=float32)
get_predictions_and_export_outputs: predictions_dict = {'generated_images': <tf.Tensor 'Reshape:0' shape=(?, 28, 28, 1) dtype=float32>}
get_predictions_and_export_outputs: export_outputs = {'predict_export_outputs': <tensorflow.python.saved_model.model_utils.export_output.PredictOutput object at 0x7f89142a5690>}
decode_example: features = {'image_raw': FixedLenFeature(shape=[], dtype=tf.string, default_value=None), 'label': FixedLenFeature(shape=[], dtype=tf.int64, default_value=None)}
decode_example: image = Tensor("DecodeRaw:0", shape=(?,), dtype=uint8)
decode_example: image = Tensor("Reshape:0", shape=(28, 28, 1), dtype=uint8)
preprocess_image: image = Tensor("sub:0", shape=(28, 28, 1), dtype=float32)
decode_example: image = Tensor("sub:0", shape=(28, 28, 1), dtype=float32)
decode_example: label = Tensor("Cast_1:0", shape=(), dtype=int32)
cgan_model: features = {'image': <tf.Tensor 'IteratorGetNext:0' shape=(?, 28, 28, 1) dtype=float32>}
cgan_model: labels = Tensor("IteratorGetNext:1", shape=(?,), dtype=int32, device=/device:CPU:0)
cgan_model: mode = eval
cgan_model: params = {'train_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord', 'eval_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord', 'output_dir': 'gs://machine-learning-1234-bucket/gan/cgan/trained_model2/', 'train_batch_size': 100, 'train_steps': 60000, 'save_summary_steps': 100, 'save_checkpoints_steps': 5000, 'keep_checkpoint_max': 10, 'eval_batch_size': 16, 'eval_steps': 10, 'start_delay_secs': 6000, 'throttle_secs': 6000, 'height': 28, 'width': 28, 'depth': 1, 'num_classes': 10, 'label_embedding_dimension': 10, 'latent_size': 512, 'generator_use_labels': True, 'generator_embed_labels': True, 'generator_concatenate_labels': True, 'generator_dense_before_concatenate': False, 'generator_hidden_units': [256, 512, 1024], 'generator_leaky_relu_alpha': 0.2, 'generator_final_activation': 'tanh', 'generator_l1_regularization_scale': 0.0, 'generator_l2_regularization_scale': 0.0, 'generator_optimizer': 'Adam', 'generator_learning_rate': 0.0002, 'generator_adam_beta1': 0.5, 'generator_adam_beta2': 0.999, 'generator_adam_epsilon': 1e-08, 'generator_clip_gradients': None, 'generator_train_steps': 1, 'discriminator_use_labels': True, 'discriminator_embed_labels': True, 'discriminator_concatenate_labels': True, 'discriminator_dense_before_concatenate': False, 'discriminator_hidden_units': [1024, 512, 256], 'discriminator_leaky_relu_alpha': 0.2, 'discriminator_l1_regularization_scale': 0.0, 'discriminator_l2_regularization_scale': 0.0, 'discriminator_optimizer': 'Adam', 'discriminator_learning_rate': 0.0002, 'discriminator_adam_beta1': 0.5, 'discriminator_adam_beta2': 0.999, 'discriminator_adam_epsilon': 1e-08, 'discriminator_clip_gradients': None, 'discriminator_train_steps': 1, 'label_smoothing': 0.9}
cgan_model: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
Training discriminator.
get_logits_and_losses: real_images = Tensor("real_images:0", shape=(?, 784), dtype=float32)
get_logits_and_losses: Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32)
Call generator with Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32).
get_fake_images: Z = Tensor("discriminator_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
Call discriminator with fake_images = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
Call discriminator with real_images = Tensor("real_images:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("real_images:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator_1/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator_1/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator_1/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_1/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator_1/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
get_discriminator_loss: discriminator_real_loss = Tensor("discriminator_real_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_fake_loss = Tensor("discriminator_fake_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_loss = Tensor("discriminator_loss:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_reg_loss = Tensor("Const_3:0", shape=(), dtype=float32)
get_discriminator_loss: discriminator_total_loss = Tensor("discriminator_total_loss:0", shape=(), dtype=float32)
Training generator.
get_logits_and_losses: fake_labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
Call generator with fake_Z = Tensor("generator_Z:0", shape=(?, 512), dtype=float32).
get_fake_images: Z = Tensor("generator_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator_1/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator_1/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator_1/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator_1/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator_1/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
Call discriminator with fake_fake_images = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32).
get_discriminator_logits: X = Tensor("generator_1/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_discriminator_logits: labels = Tensor("fake_labels:0", shape=(?, 1), dtype=int32)
discriminator_embed_labels: label_vectors = Tensor("discriminator_2/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: label_vectors = Tensor("discriminator_2/discriminator/discriminator/label_vectors:0", shape=(?, 10), dtype=float32)
discriminator_use_labels: network = Tensor("discriminator_2/discriminator/concat_labels:0", shape=(?, 794), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_0/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_0:0", shape=(?, 1024), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/layers_dense_2/BiasAdd:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: network = Tensor("discriminator_2/leaky_relu_2:0", shape=(?, 256), dtype=float32)
get_discriminator_logits: logits = Tensor("discriminator_2/layers_dense_logits/BiasAdd:0", shape=(?, 1), dtype=float32)
get_generator_loss: generator_loss = Tensor("generator_loss:0", shape=(), dtype=float32)
get_generator_loss: generator_reg_loss = Tensor("Const_5:0", shape=(), dtype=float32)
get_generator_loss: generator_total_loss = Tensor("generator_total_loss:0", shape=(), dtype=float32)
get_fake_images: Z = Tensor("image_summary_Z:0", shape=(10, 512), dtype=float32)
get_fake_images: labels = Tensor("image_summary_fake_labels:0", shape=(10, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator_2/generator/generator/label_vectors:0", shape=(10, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator_2/generator/generator/label_vectors:0", shape=(10, 10), dtype=float32)
generator_use_labels: network = Tensor("generator_2/generator/concat_labels:0", shape=(10, 522), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_0/BiasAdd:0", shape=(10, 256), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_0:0", shape=(10, 256), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_1/BiasAdd:0", shape=(10, 512), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_1:0", shape=(10, 512), dtype=float32)
get_fake_images: network = Tensor("generator_2/layers_dense_2/BiasAdd:0", shape=(10, 1024), dtype=float32)
get_fake_images: network = Tensor("generator_2/leaky_relu_2:0", shape=(10, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator_2/layers_dense_generated_outputs/Tanh:0", shape=(10, 784), dtype=float32)
get_eval_metric_ops: discriminator_logits = Tensor("discriminator_concat_logits:0", shape=(?, 1), dtype=float32)
get_eval_metric_ops: discriminator_labels = Tensor("discriminator_concat_labels:0", shape=(?, 1), dtype=float32)
get_eval_metric_ops: discriminator_probabilities = Tensor("discriminator_probabilities:0", shape=(?, 1), dtype=float32)
get_eval_metric_ops: eval_metric_ops = {'accuracy': (<tf.Tensor 'discriminator_accuracy/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_accuracy/update_op:0' shape=() dtype=float32>), 'precision': (<tf.Tensor 'discriminator_precision/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_precision/update_op:0' shape=() dtype=float32>), 'recall': (<tf.Tensor 'discriminator_recall/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_recall/update_op:0' shape=() dtype=float32>), 'auc_roc': (<tf.Tensor 'discriminator_auc_roc/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_auc_roc/update_op:0' shape=() dtype=float32>), 'auc_pr': (<tf.Tensor 'discriminator_auc_pr/value:0' shape=() dtype=float32>, <tf.Tensor 'discriminator_auc_pr/update_op:0' shape=() dtype=float32>)}
serving_input_fn: feature_placeholders = {'Z': <tf.Tensor 'serving_input_placeholder_Z:0' shape=(?, 512) dtype=float32>, 'label': <tf.Tensor 'serving_input_placeholder_label:0' shape=(?,) dtype=int32>}
serving_input_fn: features = {'Z': <tf.Tensor 'serving_input_fn_identity_placeholder_Z:0' shape=(?, 512) dtype=float32>, 'label': <tf.Tensor 'serving_input_fn_identity_placeholder_label:0' shape=(?,) dtype=int32>}
cgan_model: features = {'Z': <tf.Tensor 'serving_input_fn_identity_placeholder_Z:0' shape=(?, 512) dtype=float32>, 'label': <tf.Tensor 'serving_input_fn_identity_placeholder_label:0' shape=(?,) dtype=int32>}
cgan_model: labels = None
cgan_model: mode = infer
cgan_model: params = {'train_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/train*.tfrecord', 'eval_file_pattern': 'gs://machine-learning-1234-bucket/gan/data/mnist/test*.tfrecord', 'output_dir': 'gs://machine-learning-1234-bucket/gan/cgan/trained_model2/', 'train_batch_size': 100, 'train_steps': 60000, 'save_summary_steps': 100, 'save_checkpoints_steps': 5000, 'keep_checkpoint_max': 10, 'eval_batch_size': 16, 'eval_steps': 10, 'start_delay_secs': 6000, 'throttle_secs': 6000, 'height': 28, 'width': 28, 'depth': 1, 'num_classes': 10, 'label_embedding_dimension': 10, 'latent_size': 512, 'generator_use_labels': True, 'generator_embed_labels': True, 'generator_concatenate_labels': True, 'generator_dense_before_concatenate': False, 'generator_hidden_units': [256, 512, 1024], 'generator_leaky_relu_alpha': 0.2, 'generator_final_activation': 'tanh', 'generator_l1_regularization_scale': 0.0, 'generator_l2_regularization_scale': 0.0, 'generator_optimizer': 'Adam', 'generator_learning_rate': 0.0002, 'generator_adam_beta1': 0.5, 'generator_adam_beta2': 0.999, 'generator_adam_epsilon': 1e-08, 'generator_clip_gradients': None, 'generator_train_steps': 1, 'discriminator_use_labels': True, 'discriminator_embed_labels': True, 'discriminator_concatenate_labels': True, 'discriminator_dense_before_concatenate': False, 'discriminator_hidden_units': [1024, 512, 256], 'discriminator_leaky_relu_alpha': 0.2, 'discriminator_l1_regularization_scale': 0.0, 'discriminator_l2_regularization_scale': 0.0, 'discriminator_optimizer': 'Adam', 'discriminator_learning_rate': 0.0002, 'discriminator_adam_beta1': 0.5, 'discriminator_adam_beta2': 0.999, 'discriminator_adam_epsilon': 1e-08, 'discriminator_clip_gradients': None, 'discriminator_train_steps': 1, 'label_smoothing': 0.9}
get_predictions_and_export_outputs: Z = Tensor("serving_input_fn_identity_placeholder_Z:0", shape=(?, 512), dtype=float32)
get_predictions_and_export_outputs: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
get_fake_images: Z = Tensor("serving_input_fn_identity_placeholder_Z:0", shape=(?, 512), dtype=float32)
get_fake_images: labels = Tensor("ExpandDims:0", shape=(?, 1), dtype=int32)
generator_embed_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: label_vectors = Tensor("generator/generator/generator/label_vectors:0", shape=(?, 10), dtype=float32)
generator_use_labels: network = Tensor("generator/generator/concat_labels:0", shape=(?, 522), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_0/BiasAdd:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_0:0", shape=(?, 256), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_1/BiasAdd:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_1:0", shape=(?, 512), dtype=float32)
get_fake_images: network = Tensor("generator/layers_dense_2/BiasAdd:0", shape=(?, 1024), dtype=float32)
get_fake_images: network = Tensor("generator/leaky_relu_2:0", shape=(?, 1024), dtype=float32)
get_fake_images: generated_outputs = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_predictions_and_export_outputs: fake_images = Tensor("generator/layers_dense_generated_outputs/Tanh:0", shape=(?, 784), dtype=float32)
get_predictions_and_export_outputs: generated_images = Tensor("Reshape:0", shape=(?, 28, 28, 1), dtype=float32)
get_predictions_and_export_outputs: predictions_dict = {'generated_images': <tf.Tensor 'Reshape:0' shape=(?, 28, 28, 1) dtype=float32>}
get_predictions_and_export_outputs: export_outputs = {'predict_export_outputs': <tensorflow.python.saved_model.model_utils.export_output.PredictOutput object at 0x7f88f478ded0>}
###Markdown
Prediction
###Code
import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
!gsutil ls gs://machine-learning-1234-bucket/gan/cgan/trained_model/export/exporter
predict_fn = tf.contrib.predictor.from_saved_model(
"gs://machine-learning-1234-bucket/gan/cgan/trained_model/export/exporter/1592753841"
)
predictions = predict_fn(
{
"Z": np.random.normal(size=(num_classes, 100)),
"label": np.arange(num_classes)
}
)
print(list(predictions.keys()))
###Output
['generated_images']
###Markdown
Convert image back to the original scale.
###Code
generated_images = np.clip(
a=((predictions["generated_images"] + 1.0) * (255. / 2)).astype(np.int32),
a_min=0,
a_max=255
)
print(generated_images.shape)
def plot_images(images):
"""Plots images.
Args:
images: np.array, array of images of
[num_images, height, width, depth].
"""
num_images = len(images)
plt.figure(figsize=(20, 20))
for i in range(num_images):
image = images[i]
plt.subplot(1, num_images, i + 1)
plt.xticks([])
plt.yticks([])
plt.grid(False)
plt.imshow(
image.reshape(image.shape[:-1]),
cmap="gray_r"
)
plt.show()
plot_images(generated_images)
###Output
_____no_output_____ |
Preprocessing/SpacyCleaning-sentence.ipynb | ###Markdown
* Mark Twain: Mark + Various* Fitzgerald: Fitzgerald + Various * Charles: Charles + all
###Code
df_charles = pd.concat(charleslist,ignore_index=True)
df_charles['Label'] = "1"
df_charles = df_charles[['Text','Label']]
df_others = pd.concat(marklist +variouslsit+ janelist +fitzgeraldlist,ignore_index=True)
df_others['Label'] = "0"
df_others = df_others[['Text','Label']]
df_charles_f = pd.concat([df_charles,df_others],ignore_index=True)
print(df_charles_f.shape[0])
df_charles_f.Label.describe()
df_mark = pd.concat(marklist,ignore_index=True)
df_mark['Label'] = "1"
df_mark = df_mark[['Text','Label']]
df_others = pd.concat(variouslsit,ignore_index=True)
df_others['Label'] = "0"
df_others = df_others[['Text','Label']]
df_mark_f = pd.concat([df_mark,df_others],ignore_index=True)
print(df_mark_f.shape[0])
df_mark_f.Label.describe()
df_fitzgerald = pd.concat(fitzgeraldlist,ignore_index=True)
df_fitzgerald['Label'] = "1"
df_fitzgerald = df_fitzgerald[['Text','Label']]
df_others = pd.concat(variouslsit,ignore_index=True)
df_others['Label'] = "0"
df_others = df_others[['Text','Label']]
df_fitzgerald_f = pd.concat([df_fitzgerald,df_others],ignore_index=True)
print(df_fitzgerald_f.shape[0])
df_fitzgerald_f.Label.describe()
# apply on the df['Text']
# retain all columns and add addtional columns spacy_words
def NER_replace(df):
df['spacy_words'] = df['Text'].apply(lambda x: list(nlp(x).ents))
for index, row in df.iterrows():
thistext = row.Text
for ent in row.spacy_words:
if ent.label_ in ['PERSON','FAC','GPE','LOC','ORG']:
thistext = thistext.replace(ent.text,ent.label_)
df.at[index, 'Text'] = thistext
return df
df_fitzgerald_f = NER_replace(df_fitzgerald_f)
#df_mark_f = NER_replace(df_mark_f)
df_charles_f = NER_replace(df_charles_f)
df_fitzgerald_f.Text[40]
for ent in df_fitzgerald_f.spacy_words[40]:
if ent.label_ in ['PERSON','FAC','GPE','LOC','ORG']:
print(ent.text, ent.start_char, ent.end_char,
ent.label_)
df_fitzgerald_f.to_csv("../Dataset_v3/deidentified/Fitzgerald_sent_di.csv")
df_charles_f.to_csv("../Dataset_v3/deidentified/Charles_sent_di.csv")
#df_mark_f.to_csv("../Dataset_v3/deidentified/Mark_sent_di.csv")
###Output
_____no_output_____ |
DAY-5/DAY-5.ipynb | ###Markdown
Question 1 :Write a Python program to find the first 20 non-even prime natural numbers.
###Code
import sympy
prime_no = list(sympy.primerange(0, 75))
for i in prime_no:
if i%2!=0:
print(i)
###Output
3
5
7
11
13
17
19
23
29
31
37
41
43
47
53
59
61
67
71
73
###Markdown
Question 2 :Write a Python program to implement 15 functions of string.
###Code
# 1. strip(): The strip() method removes any whitespace from the beginning or the end.
a = " Hello, World! "
print(a.strip())
#2. lower(): The lower() method returns the string in lower case.
a = "Hello, World!"
print(a.lower())
#3. upper(): The upper() method returns the string in upper case.
a = "Hello, World!"
print(a.upper())
#4. replace(): The replace() method replaces a string with another string.
a = "Hello, World!"
print(a.replace("W", "J"))
#5. split(): The split() method splits the string into substrings if it finds instances of the separator.
a = "Hello, World!"
print(a.split(","))
#6. isdigit():
txt = "50800"
x = txt.isdigit()
print(x)
#7. isidentifier():
txt = "Demo"
x = txt.isidentifier()
print(x)
#8. islower():
txt = "hello world!"
x = txt.islower()
print(x)
#9. isupper():
txt = "THIS IS NOW!"
x = txt.isupper()
print(x)
#10. isspace():
txt = " "
x = txt.isspace()
print(x)
#11. istitle():
txt = "Hello, And Welcome To My World!"
x = txt.istitle()
print(x)
#12. isnumeric():
txt = "565543"
x = txt.isnumeric()
print(x)
#13. isprintable():
txt = "Hello! Are you #1?"
x = txt.isprintable()
print(x)
#14. isalnum():
txt = "Company12"
x = txt.isalnum()
print(x)
#15. isalpha():
txt = "CompanyX"
x = txt.isalpha()
print(x)
#16. isdecimal():
txt = "\u0033" #unicode for 3
x = txt.isdecimal()
print(x)
#17. capitalize():
txt = "hello, and welcome to my world."
x = txt.capitalize()
print (x)
#18. casefold():
txt = "Hello, And Welcome To My World!"
x = txt.casefold()
print(x)
#19. center():
txt = "banana"
x = txt.center(20)
print(x)
#20. count():
txt = "I love apples, apple are my favorite fruit"
x = txt.count("apple")
print(x)
#21. encode():
txt = "My name is Ståle"
x = txt.encode()
print(x)
#22. endswith():
txt = "Hello, welcome to my world."
x = txt.endswith(".")
print(x)
#23. expandtabs():
txt = "H\te\tl\tl\to"
x = txt.expandtabs(2)
print(x)
#24. find():
txt = "Hello, welcome to my world."
x = txt.find("welcome")
print(x)
#25. format():
txt = "For only {price:.2f} dollars!"
print(txt.format(price = 49))
#26. index():
txt = "Hello, welcome to my world."
x = txt.index("welcome")
print(x)
#27. rsplit():
txt = "apple, banana, cherry"
x = txt.rsplit(", ")
print(x)
#28. rstrip():
txt = " banana "
x = txt.rstrip()
print("of all fruits", x, "is my favorite")
#29. splitlines():
txt = "Thank you for the music\nWelcome to the jungle"
x = txt.splitlines()
print(x)
#30. startswith():
txt = "Hello, welcome to my world."
x = txt.startswith("Hello")
print(x)
#31. swapcase():
txt = "Hello My Name Is PETER"
x = txt.swapcase()
print(x)
#32. title():
txt = "Welcome to my world"
x = txt.title()
print(x)
#33. zfill():
txt = "50"
x = txt.zfill(10)
print(x)
#34. partition():
txt = "I could eat bananas all day"
x = txt.partition("bananas")
print(x)
#35. replace():
txt = "I like bananas"
x = txt.replace("bananas", "apples")
print(x)
#36. rfind():
txt = "Mi casa, su casa."
x = txt.rfind("casa")
print(x)
#37. rindex():
txt = "Mi casa, su casa."
x = txt.rindex("casa")
print(x)
#38. rjust():
txt = "banana"
x = txt.rjust(20)
print(x, "is my favorite fruit.")
#39. rpartition():
txt = "I could eat bananas all day, bananas are my favorite fruit"
x = txt.rpartition("bananas")
print(x)
#40. join():
myTuple = ("John", "Peter", "Vicky")
x = "#".join(myTuple)
print(x)
#41. ljust():
txt = "banana"
x = txt.ljust(20)
print(x, "is my favorite fruit.")
#42. lstrip():
txt = " banana "
x = txt.lstrip()
print("of all fruits", x, "is my favorite")
###Output
of all fruits banana is my favorite
###Markdown
Question 3:Write a Python program to check if the given string is a Palindrome or Anagram or None of them.Display the message accordingly to the user.
###Code
def is_anagram(str1, str2):
list_str1 = list(str1)
list_str1.sort()
list_str2 = list(str2)
list_str2.sort()
return (list_str1 == list_str2)
string = input("Enter first string to check whether it is a anagram: ")
string1 = input("Enter second string to check whether it is a anagram: ")
ana_check = is_anagram(string,string1)
pali_check = string==string[::-1]
if pali_check and ana_check:
print("The entered string is a Palindrome as well as an Anagram.")
elif pali_check:
print("The entered string is a Palindrome")
elif ana_check:
print("The entered string is an Anagram")
else:
print("None of them")
###Output
Enter first string to check whether it is a anagram: wasitacatisaw
Enter second string to check whether it is a anagram: acatisawwasit
The entered string is a Palindrome as well as an Anagram.
###Markdown
Question 4:Write a Python's user defined function that removes all the additional characters from the stringand converts it finally to lower case using built-in lower(). eg: If the string is "Dr. Darshan Ingle@AI-ML Trainer", then the output be "drdarshaningleaimltrainer".
###Code
def replace_char(z):
invalid_list = ['!', '@', '#', '$', '%', '^', '&', '*', '(', ')', '-', '_', '+', ',', '.', '?', ':', ';', " "]
for i in invalid_list:
z = z.replace(i,'')
z = z.lower()
print("After removing all special characters from the string!")
print("Resultant String is: ",z)
string = "Dr. Darshan Ingle @AI-ML Trainer."
print("The String is: ", string)
replace_char(string)
###Output
The String is: Dr. Darshan Ingle @AI-ML Trainer.
After removing all special characters from the string!
Resultant String is: drdarshaningleaimltrainer
|
TD-Learning/TD learning.ipynb | ###Markdown
引入CliffWalking-v0环境，及环境初探
###Code
import gym
env = gym.make('CliffWalking-v0') # 经典悬崖寻路环境
print('状态空间：',env.observation_space)
print('动作空间：',env.action_space)
print('状态35执行[⬆️-0,➡️-1,⬇️-2,⬅️-3]动作后的四元组，分别表示(prob ,next_state, reward, down):')
env.P[35]
env.reset()
env.render()
###Output
o o o o o o o o o o o o
o o o o o o o o o o o o
o o o o o o o o o o o o
x C C C C C C C C C C T
###Markdown
1. Solving CliffWalking-v0 via Sarsa
###Code
import numpy as np
import random
def select_action_behavior_policy(action_value_set, epsilon):
'''使用epsilon-greedy采样action'''
prob = random.random()
if prob > epsilon: action = np.argmax(action_value_set)
else: action = random.randint(0,3)
return action
def Sarsa(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9, Q_table=None):
'''
Sarsa算法,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q表
if Q_table is None: Q = np.zeros( (env.nS, env.nA), dtype=np.float )
else: Q = Q_table # 方便后续作图，不做图的话可以忽略本行和Q_table形参
sum_reward = 0
for num in range(num_of_episode):
# sum_reward = 0
state = env.reset() # Init S
# 2.通过behavior policy采样初始state下的action
action = select_action_behavior_policy(Q[state], epsilon)
while True:
# 3.执行action并观察R和next state
next_state, reward, done, info = env.step(action)
# 4.再次通过behavior policy采样next_action
next_action = select_action_behavior_policy(Q[next_state], epsilon)
# 5.更新Q(S,A),使用下一个状态-动作二元组更新
Q[state][action] += alpha * (reward + gamma*Q[next_state][next_action] - Q[state][action])
sum_reward += reward
if done: break
state, action = next_state, next_action
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
# if num % 20 == 0: print("Episode: {}, Score: {}".format(num, sum_reward))
return Q, sum_reward
def get_optimal_policy(env, Q):
'''从Q表中得到最优策略'''
nS = Q.shape[0]
policy = np.full(nS, 0) # 初始化为全0元素
for state in range(nS):
action = np.argmax( Q[state] )
next_state = env.P[state][action][0][1]
policy[state] = action
return policy
def print_policy(policy):
print("\n【Sarsa Optimal Policy】：")
for i, p in enumerate(policy):
if i % 12 == 0 and i != 0: print()
if p == 0: print('⬆️', end=' ')
elif p == 1: print('➡️', end=' ')
elif p == 2: print('⬇️', end=' ')
elif p == 3: print('⬅️', end=' ')
env = gym.make('CliffWalking-v0') # 经典悬崖寻路环境
# Q-learning学习出Q表
Q_table, _ = Sarsa(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=2000, gamma=0.9)
# 通过训练好的Q表获取最优policy
policy = get_optimal_policy(env, Q_table)
# 打印最优policy
print_policy(policy)
###Output
【Sarsa Optimal Policy】：
➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ⬇️
⬆️ ⬆️ ➡️ ⬅️ ➡️ ⬆️ ➡️ ➡️ ➡️ ➡️ ➡️ ⬇️
⬆️ ➡️ ⬆️ ⬆️ ⬆️ ⬆️ ➡️ ⬆️ ⬆️ ⬆️ ➡️ ⬇️
⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️
###Markdown
2. Solving CliffWalking-v0 via Q-learning
###Code
import numpy as np
import random
def select_action_behavior_policy(action_value_set, epsilon):
'''使用epsilon-greedy采样action'''
prob = random.random()
if prob > epsilon: action = np.argmax(action_value_set)
else: action = random.randint(0,3)
return action
def Q_learning(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9, Q_table=None):
'''
Q学习算法,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q表
if Q_table is None: Q = np.zeros( (env.nS, env.nA), dtype=np.float )
else: Q = Q_table
sum_reward = 0
for num in range(num_of_episode):
# sum_reward = 0
state = env.reset() # Init S
while True:
# 2.通过behavior policy采样action
action = select_action_behavior_policy(Q[state], epsilon)
# 3.执行action并观察R和next state
next_state, reward, done, info = env.step(action)
# 4.更新Q(S,A),使用max操作更新
Q[state][action] += alpha * (reward + gamma*max( Q[next_state] ) - Q[state][action])
sum_reward += reward
if done: break
state = next_state
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
# if num % 20 == 0: print("Episode: {}, Score: {}".format(num, sum_reward))
return Q, sum_reward
def get_optimal_policy(env, Q):
'''从Q表中得到最优策略'''
nS = Q.shape[0]
policy = np.full(nS, 0) # 初始化为全0元素
for state in range(nS):
action = np.argmax( Q[state] )
next_state = env.P[state][action][0][1]
policy[state] = action
return policy
def print_policy(policy):
print("\n【Q-learning Optimal Policy】：")
for i, p in enumerate(policy):
if i % 12 == 0 and i != 0: print()
if p == 0: print('⬆️', end=' ')
elif p == 1: print('➡️', end=' ')
elif p == 2: print('⬇️', end=' ')
elif p == 3: print('⬅️', end=' ')
env = gym.make('CliffWalking-v0') # 经典悬崖寻路环境
# Q-learning学习出Q表
Q_table, _ = Q_learning(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9)
# 通过训练好的Q表获取最优policy
policy = get_optimal_policy(env, Q_table)
# 打印最优policy
print_policy(policy)
###Output
【Q-learning Optimal Policy】：
⬆️ ➡️ ⬅️ ⬅️ ⬇️ ➡️ ➡️ ➡️ ➡️ ➡️ ⬅️ ⬇️
⬇️ ⬇️ ➡️ ➡️ ➡️ ⬇️ ⬇️ ➡️ ➡️ ➡️ ➡️ ⬇️
➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ➡️ ⬇️
⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️ ⬆️
###Markdown
- 两种算法Average reward对比
###Code
from tqdm import tqdm
import matplotlib
matplotlib.use('nbAgg')
import matplotlib.pyplot as plt
import matplotlib as mpl
def plot_Sarsa_and_Q_learning(env, episodes=500, runs=50):
'''绘图Q-learning'''
Sarsa_table, Q_learning_table, Sarsa_average_reward, Q_learning_average_reward = None, None, [], []
for i in tqdm(range(episodes)):
Sarsa_table, Sarsa_sum_reward = Sarsa(env, alpha=0.2, epsilon_scope=[0.1,0.1,0.99], num_of_episode=runs, gamma=0.9, Q_table=Sarsa_table)
Q_learning_table, Q_learning_sum_reward = Q_learning(env, alpha=0.2, epsilon_scope=[0.1,0.1,0.99], num_of_episode=runs, gamma=0.9, Q_table=Q_learning_table)
Sarsa_average_reward.append(Sarsa_sum_reward/runs)
Q_learning_average_reward.append(Q_learning_sum_reward/runs)
plt.plot(Q_learning_average_reward, label='Q-Learning')
plt.plot(Sarsa_average_reward, label='Sarsa')
plt.xlabel('Episodes')
plt.ylabel('Average of rewards during episode')
plt.ylim([-120, 0])
plt.legend()
plt.show()
plt.close()
env.reset()
plot_Sarsa_and_Q_learning(env, episodes=500, runs=60)
###Output
100%|██████████| 500/500 [00:15<00:00, 34.04it/s]
###Markdown
- 两种算法耗时对比
###Code
from timeit import repeat
consum_time = repeat( lambda:Q_learning(env, 0.2, [0.2,0.05,0.99], 1000, gamma=0.9), number=1, repeat=3 )
print('\nQ-learning在CliffWalking-v0任务上的最短耗时：%.4f秒' % ( min(consum_time) ) )
consum_time = repeat( lambda:Sarsa(env, 0.2, [0.2,0.05,0.99], 1000, gamma=0.9), number=1, repeat=3 )
print('\nSarsa在CliffWalking-v0任务上的最短耗时：%.4f秒' % ( min(consum_time) ) )
###Output
Q-learning在CliffWalking-v0任务上的最短耗时：0.3044秒
Sarsa在CliffWalking-v0任务上的最短耗时：0.2843秒
###Markdown
3. Double Q-learning - 构建环境
###Code
import numpy as np
class Env():
'''构造一个环境类'''
def __init__(self, mu, sigma, nB):
self.mu = mu
self.sigma = sigma
self.STATE_A = self.left = 0
self.STATE_B = self.right = 1
self.Terminal = 2
self.nS = 3 # 加上Terminal即3个状态
self.nA = 2
self.nB = nB # 状态B的动作数
self.state = self.STATE_A
def reset(self):
self.state = self.STATE_A
return self.state
def step(self, action):
# A--left
if self.state == self.STATE_A and action == self.left:
self.state = self.STATE_B
return self.state, 0, False # next_state, reward, done
# A--right
elif self.state == self.STATE_A and action == self.right:
self.state = self.Terminal
return self.state, 0, True
# B--all_actions
elif self.state == self.STATE_B:
self.state = self.Terminal
reward = random.normalvariate(self.mu, self.sigma)
return self.state, reward, True
###Output
_____no_output_____
###Markdown
- 初始化Q表+采样动作函数定义
###Code
import numpy as np
import random
def init_Q_table(env):
'''初始化Q表'''
Q = {env.STATE_A:{action:0 for action in range(env.nA)},
env.STATE_B:{action:0 for action in range(env.nB)},
env.Terminal:{action:0 for action in range(env.nA)}}
return Q
def select_action_behavior_policy(action_value_dict, epsilon):
'''使用epsilon-greedy采样action'''
if random.random() > epsilon:
max_keys = [key for key, value in action_value_dict.items() if value == max( action_value_dict.values() )]
action = random.choice(max_keys)
else:
# 从Q字典对应state中随机选取1个动作,由于返回list,因此通过[0]获取元素
action = random.sample(action_value_dict.keys(), 1)[0]
return action
###Output
_____no_output_____
###Markdown
- Q-learning算法实现
###Code
def Q_learning(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9):
'''
Q学习算法,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q表
Q = init_Q_table(env)
for num in range(num_of_episode):
state = env.reset() # Init S
while True:
# 2.通过behavior policy采样action
action = select_action_behavior_policy(Q[state], epsilon)
# 3.执行action并观察R和next state
next_state, reward, done = env.step(action)
# 4.更新Q(S,A),使用max操作更新
Q[state][action] += alpha * (reward + gamma*max( Q[next_state].values() ) - Q[state][action])
if done: break
state = next_state
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
return Q
env = Env(-0.1, 1, 10)
# Q-learning学习出Q表
Q_table = Q_learning(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=300, gamma=0.9)
Q_table
###Output
_____no_output_____
###Markdown
- Double Q-learning算法实现
###Code
def get_Q1_add_Q2(Q1_state_dict, Q2_state_dict):
'''返回Q1[state]+Q2[state]'''
return {action: Q1_value + Q2_state_dict[action] for action, Q1_value in Q1_state_dict.items()}
def double_Q_learning(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9):
'''
双Q学习算法,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q1表和Q2表
Q1 = init_Q_table(env)
Q2 = init_Q_table(env)
for num in range(num_of_episode):
state = env.reset() # Init S
while True:
# 2.通过behavior policy采样action
add_Q1_Q2_state = get_Q1_add_Q2(Q1[state], Q1[state])
action = select_action_behavior_policy(add_Q1_Q2_state, epsilon)
# 3.执行action并观察R和next state
next_state, reward, done = env.step(action)
# 4.更新Q(S,A),使用max操作更新
if random.random() >= 0.5:
# 从Q1表中的下一步state找出状态价值最高对应的action视为Q1[state]的最优动作
A1 = random.choice( [action for action, value in Q1[next_state].items() if value == max( Q1[next_state].values() )] )
# 将Q1[state]得到的最优动作A1代入到Q2[state][A1]中的值作为Q1[state]的更新
Q1[state][action] += alpha * (reward + gamma*Q2[next_state][A1] - Q1[state][action])
else:
A2 = random.choice( [action for action, value in Q2[next_state].items() if value == max( Q2[next_state].values() )] )
Q2[state][action] += alpha * (reward + gamma*Q1[next_state][A2] - Q2[state][action])
if done: break
state = next_state
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
return Q1
env = Env(-0.1, 1, 10)
# Q-learning学习出Q表
Q_table = double_Q_learning(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=300, gamma=0.9)
Q_table
###Output
_____no_output_____
###Markdown
- Action Distribution思想实现
###Code
def Action_Distribution(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9):
'''
按当前state的动作分布选择动作,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q表
Q = init_Q_table(env)
for num in range(num_of_episode):
state = env.reset() # Init S
while True:
# 2.通过behavior policy采样action
action = select_action_behavior_policy(Q[state], epsilon)
# 3.执行action并观察R和next state
next_state, reward, done = env.step(action)
# 4.更新Q(S,A),使用max操作更新
Q[state][action] += alpha * (reward + gamma*random.choice(list( Q[next_state].values() )) - Q[state][action])
if done: break
state = next_state
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
return Q
env = Env(-0.1, 1, 10)
# Q-learning学习出Q表
Q_table = Action_Distribution(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=300, gamma=0.9)
Q_table
###Output
_____no_output_____
###Markdown
- Expected Sarsa算法实现
###Code
def Expected_Sarsa(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9):
'''
期望Sarsa算法,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q表
Q = init_Q_table(env)
for num in range(num_of_episode):
state = env.reset() # Init S
while True:
# 2.通过behavior policy采样action
action = select_action_behavior_policy(Q[state], epsilon)
# 3.执行action并观察R和next state
next_state, reward, done = env.step(action)
# 4.更新Q(S,A),使用max操作更新
Q[state][action] += alpha * (reward + gamma*sum( Q[next_state].values() ) / len(Q[next_state]) - Q[state][action])
if done: break
state = next_state
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
return Q
env = Env(-0.1, 1, 10)
# Q-learning学习出Q表
Q_table = Expected_Sarsa(env, alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=300, gamma=0.9)
Q_table
###Output
_____no_output_____
###Markdown
TD-Learing Summary
###Code
def TD_learning(env, method='Q-Learning', alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=1000, gamma=0.9):
'''
TD学习算法,返回Q表和估计的最优策略
其中epsilon_scope由高到低衰减,从左到右分别是[最高值,最低值,衰减因子]
'''
epsilon = epsilon_scope[0]
# 1. 初始化Q1表和Q2表
Q = init_Q_table(env)
if method == 'Double-Q':
Q2 = init_Q_table(env)
bool_A_left = np.zeros(num_of_episode)
Aleft_Q_values = []
B_max_Q_values = []
for num in range(num_of_episode):
state = env.reset() # Init S
while True:
# 2.通过behavior policy采样action
if method == 'Double-Q':
add_Q1_Q2_state = {action: Q1_value + Q2[state][action] for action, Q1_value in Q[state].items()}
action = select_action_behavior_policy(add_Q1_Q2_state, epsilon)
else: action = select_action_behavior_policy(Q[state], epsilon)
if state == env.STATE_A and action == env.left:
bool_A_left[int(num)] += 1
# 3.执行action并观察R和next state
next_state, reward, done = env.step(action)
# 4.更新Q(S,A),使用max操作更新
if method == 'Q-Learning':
Q[state][action] += alpha * (reward + gamma*max( Q[next_state].values() ) - Q[state][action])
elif method == 'Expected_Sarsa':
Q[state][action] += alpha * (reward + gamma*sum( Q[next_state].values() ) / len(Q[next_state]) - Q[state][action])
elif method == 'Action_Distribution':
Q[state][action] += alpha * (reward + gamma*random.choice(list( Q[next_state].values() )) - Q[state][action])
elif method == 'Double-Q':
if random.random() >= 0.5:
# 从Q1表中的下一步state找出状态价值最高对应的action视为Q1[state]的最优动作
A1 = random.choice( [action for action, value in Q[next_state].items() if value == max( Q[next_state].values() )] )
# 将Q1[state]得到的最优动作A1代入到Q2[state][A1]中的值作为Q1[state]的更新
Q[state][action] += alpha * (reward + gamma*Q2[next_state][A1] - Q[state][action])
else:
A2 = random.choice( [action for action, value in Q2[next_state].items() if value == max( Q2[next_state].values() )] )
Q2[state][action] += alpha * (reward + gamma*Q[next_state][A2] - Q2[state][action])
if done: break
state = next_state
Aleft_Q_values.append(Q[env.STATE_A][env.left])
B_max_Q_values.append(max(Q[env.STATE_B].values()))
# 对epsilon进行衰减
if epsilon >= epsilon_scope[1]: epsilon *= epsilon_scope[2]
# if num % 20 == 0: print("Episode: {}, Score: {}".format(num, sum_reward))
return Q, bool_A_left, Aleft_Q_values, B_max_Q_values
# method = ['Q-Learning', 'Expected_Sarsa', 'Action_Distribution', 'Double-Q']
env = Env(-0.1, 1, 10)
# Q-learning学习出Q表
Q_table, _, _, _ = TD_learning(env, method='Double-Q', alpha=0.2, epsilon_scope=[0.2,0.05,0.99], num_of_episode=300, gamma=0.9)
Q_table
def show_figure(prob_Q_A_left, prob_E_A_left, prob_AD_A_left, prob_Q2_A_left):
import matplotlib.pyplot as plt
plt.ylabel('% left actions from A')
plt.xlabel('Episodes')
x_ticks = np.arange(0,301, 20)
y_ticks = np.arange(0,1.1,0.1)
plt.xticks(x_ticks)
plt.yticks(y_ticks,['0%','10%','20%','30%','40%','50%','60%','70%','80%','90%','100%'])
plt.plot(range(300), prob_Q_A_left, '-',label='Q Learning')
plt.plot(range(300), prob_E_A_left, '-',label='Double Q-Learning')
plt.plot(range(300), prob_AD_A_left, '-',label='Action Distribution')
plt.plot(range(300), prob_Q2_A_left, '-',label='Expected Sarsa')
plt.plot(np.ones(300) * 0.05, label='Optimal')
plt.title('Comparison of the effect of 4 algorithms on Ex 6.7')
plt.legend()
plt.grid()
plt.show()
plt.close()
total_num = 1000
A_Q_lst, B_Q_lst = np.zeros( (total_num, 300) ) ,np.zeros( (total_num, 300) )
A_Q2_lst, B_Q2_lst = np.zeros( (total_num, 300) ) ,np.zeros( (total_num, 300) )
A_AD_lst, B_AD_lst = np.zeros( (total_num, 300) ) ,np.zeros( (total_num, 300) )
A_E_lst, B_E_lst = np.zeros( (total_num, 300) ) ,np.zeros( (total_num, 300) )
prob_Q_A_left = np.zeros( (total_num, 300) )
prob_Q2_A_left = np.zeros( (total_num, 300) )
prob_AD_A_left = np.zeros( (total_num, 300) )
prob_E_A_left = np.zeros( (total_num, 300) )
# 计算在STATE_A下采样动作left的概率
alpha = 0.1
start_epsilon = 0.1
gamma = 0.9
num_of_episode = 300
for num in tqdm(range(total_num)):
_, A_left1, A_Q1, B_Q1 = TD_learning(env, 'Q-Learning', alpha, epsilon_scope=[start_epsilon,0.05,1], num_of_episode=num_of_episode, gamma=gamma)
_, A_left2, A_Q2, B_Q2 = TD_learning(env, 'Double-Q', alpha, epsilon_scope=[start_epsilon,0.05,1], num_of_episode=num_of_episode, gamma=gamma)
_, A_left3, A_Q3, B_Q3 = TD_learning(env, 'Action_Distribution', alpha, epsilon_scope=[start_epsilon,0.05,1], num_of_episode=num_of_episode, gamma=gamma)
_, A_left4, A_Q4, B_Q4 = TD_learning(env, 'Expected_Sarsa', alpha, epsilon_scope=[start_epsilon,0.05,1], num_of_episode=num_of_episode, gamma=gamma)
prob_Q_A_left[int(num)] = A_left1
prob_Q2_A_left[int(num)] = A_left2
prob_AD_A_left[int(num)] = A_left3
prob_E_A_left[int(num)] = A_left4
A_Q_lst[int(num)], B_Q_lst[int(num)] = A_Q1, B_Q1
A_Q2_lst[int(num)], B_Q2_lst[int(num)] = A_Q2, B_Q2
A_AD_lst[int(num)], B_AD_lst[int(num)] = A_Q3, B_Q3
A_E_lst[int(num)], B_E_lst[int(num)] = A_Q4, B_Q4
a = prob_Q_A_left.mean(axis=0)
b = prob_Q2_A_left.mean(axis=0)
c = prob_AD_A_left.mean(axis=0)
d = prob_E_A_left.mean(axis=0)
show_figure(a, b, c, d)
###Output
100%|██████████| 1000/1000 [00:09<00:00, 107.57it/s]
|
experimental/widgets/2_Widget List.ipynb | ###Markdown
[Index](Index.ipynb) - [Back](Widget Basics.ipynb) - [Next](Output Widget.ipynb) Widget List
###Code
import ipywidgets as widgets
###Output
_____no_output_____
###Markdown
Numeric widgets There are many widgets distributed with IPython that are designed to display numeric values. Widgets exist for displaying integers and floats, both bounded and unbounded. The integer widgets share a similar naming scheme to their floating point counterparts. By replacing `Float` with `Int` in the widget name, you can find the Integer equivalent. IntSlider
###Code
widgets.IntSlider(
value=7,
min=0,
max=10,
step=1,
description='Test:',
disabled=False,
continuous_update=False,
orientation='horizontal',
readout=True,
readout_format='d'
)
###Output
_____no_output_____
###Markdown
FloatSlider
###Code
widgets.FloatSlider(
value=7.5,
min=0,
max=10.0,
step=0.1,
description='Test:',
disabled=False,
continuous_update=False,
orientation='horizontal',
readout=True,
readout_format='.1f',
)
###Output
_____no_output_____
###Markdown
Sliders can also be **displayed vertically**.
###Code
widgets.FloatSlider(
value=7.5,
min=0,
max=10.0,
step=0.1,
description='Test:',
disabled=False,
continuous_update=False,
orientation='vertical',
readout=True,
readout_format='.1f',
)
###Output
_____no_output_____
###Markdown
FloatLogSlider The `FloatLogSlider` has a log scale, which makes it easy to have a slider that covers a wide range of positive magnitudes. The `min` and `max` refer to the minimum and maximum exponents of the base, and the `value` refers to the actual value of the slider.
###Code
widgets.FloatLogSlider(
value=10,
base=10,
min=-10, # max exponent of base
max=10, # min exponent of base
step=0.2, # exponent step
description='Log Slider'
)
###Output
_____no_output_____
###Markdown
IntRangeSlider
###Code
widgets.IntRangeSlider(
value=[5, 7],
min=0,
max=10,
step=1,
description='Test:',
disabled=False,
continuous_update=False,
orientation='horizontal',
readout=True,
readout_format='d',
)
###Output
_____no_output_____
###Markdown
FloatRangeSlider
###Code
widgets.FloatRangeSlider(
value=[5, 7.5],
min=0,
max=10.0,
step=0.1,
description='Test:',
disabled=False,
continuous_update=False,
orientation='horizontal',
readout=True,
readout_format='.1f',
)
###Output
_____no_output_____
###Markdown
IntProgress
###Code
widgets.IntProgress(
value=7,
min=0,
max=10,
step=1,
description='Loading:',
bar_style='', # 'success', 'info', 'warning', 'danger' or ''
orientation='horizontal'
)
###Output
_____no_output_____
###Markdown
FloatProgress
###Code
widgets.FloatProgress(
value=7.5,
min=0,
max=10.0,
step=0.1,
description='Loading:',
bar_style='info',
orientation='horizontal'
)
###Output
_____no_output_____
###Markdown
The numerical text boxes that impose some limit on the data (range, integer-only) impose that restriction when the user presses enter. BoundedIntText
###Code
widgets.BoundedIntText(
value=7,
min=0,
max=10,
step=1,
description='Text:',
disabled=False
)
###Output
_____no_output_____
###Markdown
BoundedFloatText
###Code
widgets.BoundedFloatText(
value=7.5,
min=0,
max=10.0,
step=0.1,
description='Text:',
disabled=False
)
###Output
_____no_output_____
###Markdown
IntText
###Code
widgets.IntText(
value=7,
description='Any:',
disabled=False
)
###Output
_____no_output_____
###Markdown
FloatText
###Code
widgets.FloatText(
value=7.5,
description='Any:',
disabled=False
)
###Output
_____no_output_____
###Markdown
Boolean widgets There are three widgets that are designed to display a boolean value. ToggleButton
###Code
widgets.ToggleButton(
value=False,
description='Click me',
disabled=False,
button_style='', # 'success', 'info', 'warning', 'danger' or ''
tooltip='Description',
icon='check'
)
###Output
_____no_output_____
###Markdown
Checkbox
###Code
widgets.Checkbox(
value=False,
description='Check me',
disabled=False
)
###Output
_____no_output_____
###Markdown
ValidThe valid widget provides a read-only indicator.
###Code
widgets.Valid(
value=False,
description='Valid!',
)
###Output
_____no_output_____
###Markdown
Selection widgets There are several widgets that can be used to display single selection lists, and two that can be used to select multiple values. All inherit from the same base class. You can specify the **enumeration of selectable options by passing a list** (options are either (label, value) pairs, or simply values for which the labels are derived by calling `str`). Dropdown
###Code
widgets.Dropdown(
options=['1', '2', '3'],
value='2',
description='Number:',
disabled=False,
)
###Output
_____no_output_____
###Markdown
RadioButtons
###Code
widgets.RadioButtons(
options=['pepperoni', 'pineapple', 'anchovies'],
# value='pineapple',
description='Pizza topping:',
disabled=False
)
###Output
_____no_output_____
###Markdown
Select
###Code
widgets.Select(
options=['Linux', 'Windows', 'OSX'],
value='OSX',
# rows=10,
description='OS:',
disabled=False
)
###Output
_____no_output_____
###Markdown
SelectionSlider
###Code
widgets.SelectionSlider(
options=['scrambled', 'sunny side up', 'poached', 'over easy'],
value='sunny side up',
description='I like my eggs ...',
disabled=False,
continuous_update=False,
orientation='horizontal',
readout=True
)
###Output
_____no_output_____
###Markdown
SelectionRangeSliderThe value, index, and label keys are 2-tuples of the min and max values selected. The options must be nonempty.
###Code
import datetime
dates = [datetime.date(2015,i,1) for i in range(1,13)]
options = [(i.strftime('%b'), i) for i in dates]
widgets.SelectionRangeSlider(
options=options,
index=(0,11),
description='Months (2015)',
disabled=False
)
###Output
_____no_output_____
###Markdown
ToggleButtons
###Code
widgets.ToggleButtons(
options=['Slow', 'Regular', 'Fast'],
description='Speed:',
disabled=False,
button_style='', # 'success', 'info', 'warning', 'danger' or ''
tooltips=['Description of slow', 'Description of regular', 'Description of fast'],
# icons=['check'] * 3
)
###Output
_____no_output_____
###Markdown
SelectMultipleMultiple values can be selected with shift and/or ctrl (or command) pressed and mouse clicks or arrow keys.
###Code
widgets.SelectMultiple(
options=['Apples', 'Oranges', 'Pears'],
value=['Oranges'],
#rows=10,
description='Fruits',
disabled=False
)
###Output
_____no_output_____
###Markdown
String widgets There are several widgets that can be used to display a string value. The `Text` and `Textarea` widgets accept input. The `HTML` and `HTMLMath` widgets display a string as HTML (`HTMLMath` also renders math). The `Label` widget can be used to construct a custom control label. Text
###Code
widgets.Text(
value='Hello World',
placeholder='Type something',
description='String:',
disabled=False
)
###Output
_____no_output_____
###Markdown
Textarea
###Code
widgets.Textarea(
value='Hello World',
placeholder='Type something',
description='String:',
disabled=False
)
###Output
_____no_output_____
###Markdown
LabelThe `Label` widget is useful if you need to build a custom description next to a control using similar styling to the built-in control descriptions.
###Code
widgets.HBox([widgets.Label(value="The $m$ in $E=mc^2$:"), widgets.FloatSlider()])
###Output
_____no_output_____
###Markdown
HTML
###Code
widgets.HTML(
value="Hello <b>World</b>",
placeholder='Some HTML',
description='Some HTML',
)
###Output
_____no_output_____
###Markdown
HTML Math
###Code
widgets.HTMLMath(
value=r"Some math and <i>HTML</i>: \(x^2\) and $$\frac{x+1}{x-1}$$",
placeholder='Some HTML',
description='Some HTML',
)
###Output
_____no_output_____
###Markdown
Image
###Code
file = open("images/WidgetArch.png", "rb")
image = file.read()
widgets.Image(
value=image,
format='png',
width=300,
height=400,
)
###Output
_____no_output_____
###Markdown
Button
###Code
widgets.Button(
description='Click me',
disabled=False,
button_style='', # 'success', 'info', 'warning', 'danger' or ''
tooltip='Click me',
icon='check'
)
###Output
_____no_output_____
###Markdown
OutputThe `Output` widget can capture and display stdout, stderr and [rich output generated by IPython](http://ipython.readthedocs.io/en/stable/api/generated/IPython.display.htmlmodule-IPython.display). For detailed documentation, see the [output widget examples](https://ipywidgets.readthedocs.io/en/latest/examples/Output Widget.html). Play (Animation) widget The `Play` widget is useful to perform animations by iterating on a sequence of integers with a certain speed. The value of the slider below is linked to the player.
###Code
play = widgets.Play(
# interval=10,
value=50,
min=0,
max=100,
step=1,
description="Press play",
disabled=False
)
slider = widgets.IntSlider()
widgets.jslink((play, 'value'), (slider, 'value'))
widgets.HBox([play, slider])
###Output
_____no_output_____
###Markdown
Date pickerThe date picker widget works in Chrome, Firefox and IE Edge, but does not currently work in Safari because it does not support the HTML date input field.
###Code
widgets.DatePicker(
description='Pick a Date',
disabled=False
)
###Output
_____no_output_____
###Markdown
Color picker
###Code
widgets.ColorPicker(
concise=False,
description='Pick a color',
value='blue',
disabled=False
)
###Output
_____no_output_____
###Markdown
ControllerThe `Controller` allows a game controller to be used as an input device.
###Code
widgets.Controller(
index=0,
)
###Output
_____no_output_____
###Markdown
Container/Layout widgetsThese widgets are used to hold other widgets, called children. Each has a `children` property that may be set either when the widget is created or later. Box
###Code
items = [widgets.Label(str(i)) for i in range(4)]
widgets.Box(items)
###Output
_____no_output_____
###Markdown
HBox
###Code
items = [widgets.Label(str(i)) for i in range(4)]
widgets.HBox(items)
###Output
_____no_output_____
###Markdown
VBox
###Code
items = [widgets.Label(str(i)) for i in range(4)]
left_box = widgets.VBox([items[0], items[1]])
right_box = widgets.VBox([items[2], items[3]])
widgets.HBox([left_box, right_box])
###Output
_____no_output_____
###Markdown
Accordion
###Code
accordion = widgets.Accordion(children=[widgets.IntSlider(), widgets.Text()])
accordion.set_title(0, 'Slider')
accordion.set_title(1, 'Text')
accordion
###Output
_____no_output_____
###Markdown
TabsIn this example the children are set after the tab is created. Titles for the tabes are set in the same way they are for `Accordion`.
###Code
tab_contents = ['P0', 'P1', 'P2', 'P3', 'P4']
children = [widgets.Text(description=name) for name in tab_contents]
tab = widgets.Tab()
tab.children = children
for i in range(len(children)):
tab.set_title(i, str(i))
tab
###Output
_____no_output_____
###Markdown
Accordion and Tab use `selected_index`, not valueUnlike the rest of the widgets discussed earlier, the container widgets `Accordion` and `Tab` update their `selected_index` attribute when the user changes which accordion or tab is selected. That means that you can both see what the user is doing *and* programmatically set what the user sees by setting the value of `selected_index`.Setting `selected_index = None` closes all of the accordions or deselects all tabs. In the cells below try displaying or setting the `selected_index` of the `tab` and/or `accordion`.
###Code
tab.selected_index = 3
accordion.selected_index = None
###Output
_____no_output_____
###Markdown
Nesting tabs and accordionsTabs and accordions can be nested as deeply as you want. If you have a few minutes, try nesting a few accordions or putting an accordion inside a tab or a tab inside an accordion. The example below makes a couple of tabs with an accordion children in one of them
###Code
tab_nest = widgets.Tab()
tab_nest.children = [accordion, accordion]
tab_nest.set_title(0, 'An accordion')
tab_nest.set_title(1, 'Copy of the accordion')
tab_nest
###Output
_____no_output_____ |
docs/source/notebooks/two_knife_edges.ipynb | ###Markdown
Double knife-edge diffractionexample [Vavilov S. A., Lytaev M. S. Modeling Equation for Multiple Knife-Edge Diffraction//IEEE Transactions on Antennas and Propagation. – 2020. – Vol. 68. – Iss. 5. – pp. 3869-3877.](https://doi.org/10.1109/TAP.2019.2957085)
###Code
import os
os.chdir('../../')
from rwp.kediffraction import *
from rwp.antennas import *
from rwp.environment import *
from rwp.vis import *
###Output
_____no_output_____
###Markdown
Preparing environment
###Code
env = Troposphere(flat=True)
env.z_max = 150
env.knife_edges = [KnifeEdge(range=200, height=50), KnifeEdge(range=800, height=50)]
antenna = GaussAntenna(freq_hz=300e6, height=50, beam_width=15, eval_angle=0, polarz='H')
###Output
_____no_output_____
###Markdown
Starting calculation
###Code
kdc = KnifeEdgeDiffractionCalculator(src=antenna, env=env, max_range_m=1000)
field = kdc.calculate()
###Output
_____no_output_____
###Markdown
Visualising results
###Code
vis = FieldVisualiser(field, env=env, trans_func=lambda v: 10 * cm.log10(1e-16 + abs(v)), label='ke')
plt = vis.plot2d(min=-40, max=0)
plt.title('The intensity of the field component 10log10|u|')
plt.xlabel('Range (m)')
plt.ylabel('Height (m)')
plt.show()
plt = vis.plot_hor(51)
plt.title('The intensity of the field component 10log10|u| at the height 51 m')
plt.xlabel('Range (m)')
plt.ylabel('10log|u| (dB)')
plt.show()
###Output
_____no_output_____ |
4_channelomics/07_build_figure_supplement.ipynb | ###Markdown
Figures for supplement Histograms summarizing fits
###Code
import matplotlib as mpl
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import pickle
%matplotlib inline
import sys; sys.path.append('../')
from common import col, svg
!mkdir -p svg/
nfidxs = pickle.load(open('results/net_maf/idxs.pkl', 'rb'))
nfccs = np.asarray(pickle.load(open('results/net_maf/ccs.pkl', 'rb')))
mfidxs = pickle.load(open('results/manual_fit/idxs.pkl', 'rb'))
mfccs = np.asarray(pickle.load(open('results/manual_fit/ccs.pkl', 'rb')))
# We exclude CCs smaller than 0.9 since those models are outliers and the problem
# is with the modfiles rather than with inference/fitting.
cutoff = 0.9
nfidxs_valid = np.asarray(nfidxs)[np.argwhere(nfccs > cutoff).reshape(-1)]
nfccs_valid = nfccs[np.argwhere(nfccs > cutoff).reshape(-1)]
nfidxs_valid_sorted = nfidxs_valid[np.argsort(nfccs_valid).reshape(-1)]
nfccs_valid_sorted = nfccs_valid[np.argsort(nfccs_valid).reshape(-1)]
mfidxs_valid = np.asarray(mfidxs)[np.argwhere(mfccs > cutoff).reshape(-1)]
mfccs_valid = mfccs[np.argwhere(mfccs > cutoff).reshape(-1)]
mfidxs_valid_sorted = mfidxs_valid[np.argsort(mfccs_valid).reshape(-1)]
mfccs_valid_sorted = mfccs_valid[np.argsort(mfccs_valid).reshape(-1)]
ccs_combined = np.empty((373, 2))
ccs_combined[nfidxs_valid,0] = nfccs_valid
ccs_combined[mfidxs_valid,1] = mfccs_valid
with mpl.rc_context(fname='../.matplotlibrc'):
plt.figure(figsize=(10/2.54, 8/2.54))
plt.hist(ccs_combined, bins=np.linspace(0.9, 1.0, 11),
color=[col['SNPE'], col['MCMC']],
label=['posterior mode', 'curve fitting'],
fill=False, histtype='step', density=False, linewidth=1.5, clip_on=False)
plt.ylim([0, 350])
plt.legend(loc='upper left', frameon=False, title='Parameters via')
sns.despine(offset=10, trim=True)
plt.xticks(np.linspace(0.9, 1.0, 5))
plt.ylabel('# models')
plt.xlabel('CC between observation and prediction')
PANEL_CCS = 'fig/fig4_channelomics_supp_hists.svg'
plt.savefig(PANEL_CCS, transparent=True)
plt.close()
svg(PANEL_CCS)
###Output
_____no_output_____ |
queue_imbalance/logistic_regression/queue_imbalance-9065.ipynb | ###Markdown
Testing of queue imbalance for stock 9095Order of this notebook is as follows:1. [Data](Data)2. [Data visualization](Data-visualization)3. [Tests](Tests)4. [Conclusions](Conclusions)Goal is to implement queue imbalance predictor from [[1]](Resources).
###Code
%matplotlib inline
import warnings
import matplotlib.dates as md
import matplotlib.pyplot as plt
import seaborn as sns
from lob_data_utils import lob
from sklearn.metrics import roc_curve, roc_auc_score
warnings.filterwarnings('ignore')
###Output
_____no_output_____
###Markdown
DataMarket is open between 8-16 on every weekday. We decided to use data from only 9-15 for each day. Test and train dataFor training data we used data from 2013-09-01 - 2013-11-16:* 0916* 1001* 1016* 1101The data from 09-01 is the test data.
###Code
df, df_test = lob.load_prepared_data('9065', data_dir='../data/prepared/', length=None)
df.head()
###Output
Training set length for 9065: 10320
Testing set length for 9065: 3810
###Markdown
Data visualization
###Code
df['sum_buy_bid'].plot(label='total size of buy orders', style='--')
df['sum_sell_ask'].plot(label='total size of sell orders', style='-')
plt.title('Summed volumens for ask and bid lists')
plt.xlabel('Time')
plt.ylabel('Whole volume')
plt.legend()
df[['bid_price', 'ask_price', 'mid_price']].plot(style='.')
plt.legend()
plt.title('Prices')
plt.xlabel('Time')
plt.ylabel('Price')
sns.jointplot(x="mid_price", y="queue_imbalance", data=df.loc[:, ['mid_price', 'queue_imbalance']], kind="kde")
plt.title('Density')
plt.plot()
df['mid_price_indicator'].plot('kde')
plt.legend()
plt.xlabel('Mid price indicator')
plt.title('Mid price indicator density')
df['queue_imbalance'].plot('kde')
plt.legend()
plt.xlabel('Queue imbalance')
plt.title('Queue imbalance density')
###Output
_____no_output_____
###Markdown
TestsWe use logistic regression to predict `mid_price_indicator`. Mean square error We calculate residual $r_i$:$$ r_i = \hat{y_i} - y_i $$where $$ \hat{y}(I) = \frac{1}{1 + e^{-(x_0 + Ix_1 )}}$$Calculating mean square residual for all observations in the testing set is also useful to assess the predictive power.The predective power of null-model is 25%.
###Code
reg = lob.logistic_regression(df, 0, len(df))
probabilities = reg.predict_proba(df_test['queue_imbalance'].values.reshape(-1,1))
probabilities = [p1 for p0, p1 in probabilities]
err = ((df_test['mid_price_indicator'] - probabilities) ** 2).mean()
predictions = reg.predict(df_test['queue_imbalance'].values.reshape(-1, 1))
print('Mean square error is', err)
###Output
Mean square error is 0.24600292829391873
###Markdown
Logistic regression fit curve
###Code
plt.plot(df_test['queue_imbalance'].values,
lob.sigmoid(reg.coef_[0] * df_test['queue_imbalance'].values + reg.intercept_))
plt.title('Logistic regression fit curve')
plt.xlabel('Imbalance')
plt.ylabel('Probability')
###Output
_____no_output_____
###Markdown
ROC curveFor assessing the predectivity power we can calculate ROC score.
###Code
a, b, c = roc_curve(df_test['mid_price_indicator'], predictions)
logit_roc_auc = roc_auc_score(df_test['mid_price_indicator'], predictions)
plt.plot(a, b, label='predictions (area {})'.format(logit_roc_auc))
plt.plot([0, 1], [0, 1], color='navy', linestyle='--')
plt.xlabel('False Positive Rate')
plt.ylabel('True Positive Rate')
plt.legend()
st = 0
end = len(df)
plt.plot(df_test.index[st:end], predictions[st:end], 'ro', label='prediction')
plt.plot(df_test.index[st:end], probabilities[st:end], 'g.', label='probability')
plt.plot(df_test.index[st:end], df_test['mid_price_indicator'].values[st:end], 'b.', label='mid price')
plt.xticks(rotation=25)
plt.legend(loc=1)
plt.xlabel('Time')
plt.ylabel('Mid price prediction')
###Output
_____no_output_____ |
samples/dynamic_shapes/dynamic_shapes.ipynb | ###Markdown
Copyright 2021 The IREE Authors
###Code
#@title Licensed under the Apache License v2.0 with LLVM Exceptions.
# See https://llvm.org/LICENSE.txt for license information.
# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
###Output
_____no_output_____
###Markdown
Dynamic ShapesThis notebook1. Creates a TensorFlow program with dynamic shapes2. Imports that program into IREE's compiler3. Compiles the imported program to an IREE VM bytecode module4. Tests running the compiled VM module using IREE's runtime5. Downloads compilation artifacts for use with the native (C API) sample application
###Code
#@title General setup
import os
import tempfile
ARTIFACTS_DIR = os.path.join(tempfile.gettempdir(), "iree", "colab_artifacts")
os.makedirs(ARTIFACTS_DIR, exist_ok=True)
print(f"Using artifacts directory '{ARTIFACTS_DIR}'")
###Output
Using artifacts directory '/tmp/iree/colab_artifacts'
###Markdown
Create a program using TensorFlow and import it into IREENOTE: as in other domains, providing more information to a compiler allows itto generate more efficient code. As a general rule, the slowest varyingdimensions of program data like batch index or timestep are safer to treat asdynamic than faster varying dimensions like image x/y/channel. See[this paper](https://arxiv.org/pdf/2006.03031.pdf) for a discussion of thechallenges imposed by dynamic shapes and one project's approach to addressingthem.
###Code
#@title Define a sample TensorFlow module using dynamic shapes
import tensorflow as tf
class DynamicShapesModule(tf.Module):
# reduce_sum_1d (dynamic input size, static output size)
# e.g. [1, 2, 3] -> 6
@tf.function(input_signature=[tf.TensorSpec([None], tf.int32)])
def reduce_sum_1d(self, values):
return tf.math.reduce_sum(values)
# reduce_sum_2d (partially dynamic input size, static output size)
# e.g. [[1, 2, 3], [10, 20, 30]] -> [11, 22, 33]
@tf.function(input_signature=[tf.TensorSpec([None, 3], tf.int32)])
def reduce_sum_2d(self, values):
return tf.math.reduce_sum(values, 0)
# add_one (dynamic input size, dynamic output size)
# e.g. [1, 2, 3] -> [2, 3, 4]
@tf.function(input_signature=[tf.TensorSpec([None], tf.int32)])
def add_one(self, values):
return tf.math.add(values, tf.constant(1, dtype=tf.int32))
%%capture
!python -m pip install iree-compiler iree-tools-tf -f https://github.com/google/iree/releases
#@title Import the TensorFlow program into IREE as MLIR
from IPython.display import clear_output
from iree.compiler import tf as tfc
compiler_module = tfc.compile_module(
DynamicShapesModule(), import_only=True,
output_mlir_debuginfo=False)
clear_output() # Skip over TensorFlow's output.
# Print the imported MLIR to see how the compiler views this program.
print("Dynamic Shapes MLIR:\n```\n%s```\n" % compiler_module.decode("utf-8"))
# Save the imported MLIR to disk.
imported_mlir_path = os.path.join(ARTIFACTS_DIR, "dynamic_shapes.mlir")
with open(imported_mlir_path, "wt") as output_file:
output_file.write(compiler_module.decode("utf-8"))
print(f"Wrote MLIR to path '{imported_mlir_path}'")
###Output
Dynamic Shapes MLIR:
```
"builtin.module"() ( {
"builtin.func"() ( {
^bb0(%arg0: !iree_input.buffer_view): // no predecessors
%0 = "iree_input.cast.buffer_view_to_tensor"(%arg0) : (!iree_input.buffer_view) -> tensor<?xi32>
%1 = "std.call"(%0) {callee = @__inference_add_one_70} : (tensor<?xi32>) -> tensor<?xi32>
%2 = "iree_input.cast.tensor_to_buffer_view"(%1) : (tensor<?xi32>) -> !iree_input.buffer_view
"std.return"(%2) : (!iree_input.buffer_view) -> ()
}) {iree.abi = "{\22a\22:[[\22ndarray\22,\22i32\22,1,null]],\22r\22:[[\22ndarray\22,\22i32\22,1,null]],\22v\22:1}", sym_name = "add_one", type = (!iree_input.buffer_view) -> !iree_input.buffer_view} : () -> ()
"builtin.func"() ( {
^bb0(%arg0: tensor<?xi32>): // no predecessors
%0 = "mhlo.constant"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
%1 = "chlo.broadcast_add"(%arg0, %0) {broadcast_dimensions = dense<> : tensor<0xi64>} : (tensor<?xi32>, tensor<i32>) -> tensor<?xi32>
"std.return"(%1) : (tensor<?xi32>) -> ()
}) {arg_attrs = [{tf._user_specified_name = "values"}], sym_name = "__inference_add_one_70", sym_visibility = "private", tf._construction_context = "kEagerRuntime", tf._input_shapes = [#tf_type.shape<?>], type = (tensor<?xi32>) -> tensor<?xi32>} : () -> ()
"builtin.func"() ( {
^bb0(%arg0: !iree_input.buffer_view): // no predecessors
%0 = "mhlo.constant"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
%1 = "iree_input.cast.buffer_view_to_tensor"(%arg0) : (!iree_input.buffer_view) -> tensor<?xi32>
%2 = "mhlo.reduce"(%1, %0) ( {
^bb0(%arg1: tensor<i32>, %arg2: tensor<i32>): // no predecessors
%4 = "mhlo.add"(%arg1, %arg2) : (tensor<i32>, tensor<i32>) -> tensor<i32>
"mhlo.return"(%4) : (tensor<i32>) -> ()
}) {dimensions = dense<0> : tensor<1xi64>} : (tensor<?xi32>, tensor<i32>) -> tensor<i32>
%3 = "iree_input.cast.tensor_to_buffer_view"(%2) : (tensor<i32>) -> !iree_input.buffer_view
"std.return"(%3) : (!iree_input.buffer_view) -> ()
}) {iree.abi = "{\22a\22:[[\22ndarray\22,\22i32\22,1,null]],\22r\22:[[\22ndarray\22,\22i32\22,0]],\22v\22:1}", sym_name = "reduce_sum_1d", type = (!iree_input.buffer_view) -> !iree_input.buffer_view} : () -> ()
"builtin.func"() ( {
^bb0(%arg0: !iree_input.buffer_view): // no predecessors
%0 = "mhlo.constant"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
%1 = "iree_input.cast.buffer_view_to_tensor"(%arg0) : (!iree_input.buffer_view) -> tensor<?x3xi32>
%2 = "mhlo.reduce"(%1, %0) ( {
^bb0(%arg1: tensor<i32>, %arg2: tensor<i32>): // no predecessors
%4 = "mhlo.add"(%arg1, %arg2) : (tensor<i32>, tensor<i32>) -> tensor<i32>
"mhlo.return"(%4) : (tensor<i32>) -> ()
}) {dimensions = dense<0> : tensor<1xi64>} : (tensor<?x3xi32>, tensor<i32>) -> tensor<3xi32>
%3 = "iree_input.cast.tensor_to_buffer_view"(%2) : (tensor<3xi32>) -> !iree_input.buffer_view
"std.return"(%3) : (!iree_input.buffer_view) -> ()
}) {iree.abi = "{\22a\22:[[\22ndarray\22,\22i32\22,2,null,3]],\22r\22:[[\22ndarray\22,\22i32\22,1,3]],\22v\22:1}", sym_name = "reduce_sum_2d", type = (!iree_input.buffer_view) -> !iree_input.buffer_view} : () -> ()
}) : () -> ()
```
Wrote MLIR to path '/tmp/iree/colab_artifacts/dynamic_shapes.mlir'
###Markdown
Test the imported program_Note: you can stop after each step and use intermediate outputs with other tools outside of Colab.__See the [README](https://github.com/google/iree/tree/main/iree/samples/dynamic_shapesinstructions) for more details and example command line instructions._* _The "imported MLIR" can be used by IREE's generic compiler tools_* _The "flatbuffer blob" can be saved and used by runtime applications__The specific point at which you switch from Python to native tools will depend on your project._
###Code
%%capture
!python -m pip install iree-compiler -f https://github.com/google/iree/releases
#@title Compile the imported MLIR further into an IREE VM bytecode module
from iree.compiler import compile_str
# Note: we'll use the dylib-llvm-aot backend since it has the best support
# for dynamic shapes among our compiler targets.
flatbuffer_blob = compile_str(compiler_module, target_backends=["dylib-llvm-aot"], input_type="mhlo")
# Note: the dylib-llvm-aot target produces platform-specific code. Since you
# may need to recompile it yourself using the appropriate
# `-iree-llvm-target-triple` flag, we skip saving it to disk and downloading it.
%%capture
!python -m pip install iree-runtime -f https://github.com/google/iree/releases
#@title Test running the compiled VM module using IREE's runtime
from iree import runtime as ireert
vm_module = ireert.VmModule.from_flatbuffer(flatbuffer_blob)
config = ireert.Config("dylib")
ctx = ireert.SystemContext(config=config)
ctx.add_vm_module(vm_module)
import numpy as np
# Our @tf.functions are accessible by name on the module named 'module'
dynamic_shapes_program = ctx.modules.module
print(dynamic_shapes_program.reduce_sum_1d(np.array([1, 10, 100], dtype=np.int32)))
print(dynamic_shapes_program.reduce_sum_2d(np.array([[1, 2, 3], [10, 20, 30]], dtype=np.int32)))
print(dynamic_shapes_program.reduce_sum_2d(np.array([[1, 2, 3], [10, 20, 30], [100, 200, 300]], dtype=np.int32)))
print(dynamic_shapes_program.add_one(np.array([1, 10, 100], dtype=np.int32)))
###Output
111
[11 22 33]
[111 222 333]
[ 2 11 101]
###Markdown
Download compilation artifacts
###Code
ARTIFACTS_ZIP = "/tmp/dynamic_shapes_colab_artifacts.zip"
print(f"Zipping '{ARTIFACTS_DIR}' to '{ARTIFACTS_ZIP}' for download...")
!cd {ARTIFACTS_DIR} && zip -r {ARTIFACTS_ZIP} .
# Note: you can also download files using Colab's file explorer
try:
from google.colab import files
print("Downloading the artifacts zip file...")
files.download(ARTIFACTS_ZIP)
except ImportError:
print("Missing google_colab Python package, can't download files")
###Output
Zipping '/tmp/iree/colab_artifacts' to '/tmp/dynamic_shapes_colab_artifacts.zip' for download...
adding: dynamic_shapes.mlir (deflated 80%)
Downloading the artifacts zip file...
###Markdown
Copyright 2021 The IREE Authors
###Code
#@title Licensed under the Apache License v2.0 with LLVM Exceptions.
# See https://llvm.org/LICENSE.txt for license information.
# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
###Output
_____no_output_____
###Markdown
Dynamic ShapesThis notebook1. Creates a TensorFlow program with dynamic shapes2. Imports that program into IREE's compiler3. Compiles the imported program to an IREE VM bytecode module4. Tests running the compiled VM module using IREE's runtime5. Downloads compilation artifacts for use with the native (C API) sample application
###Code
#@title General setup
import os
import tempfile
ARTIFACTS_DIR = os.path.join(tempfile.gettempdir(), "iree", "colab_artifacts")
os.makedirs(ARTIFACTS_DIR, exist_ok=True)
print(f"Using artifacts directory '{ARTIFACTS_DIR}'")
###Output
Using artifacts directory '/tmp/iree/colab_artifacts'
###Markdown
Create a program using TensorFlow and import it into IREENOTE: as in other domains, providing more information to a compiler allows itto generate more efficient code. As a general rule, the slowest varyingdimensions of program data like batch index or timestep are safer to treat asdynamic than faster varying dimensions like image x/y/channel. See[this paper](https://arxiv.org/pdf/2006.03031.pdf) for a discussion of thechallenges imposed by dynamic shapes and one project's approach to addressingthem.
###Code
#@title Define a sample TensorFlow module using dynamic shapes
import tensorflow as tf
class DynamicShapesModule(tf.Module):
# reduce_sum_1d (dynamic input size, static output size)
# e.g. [1, 2, 3] -> 6
@tf.function(input_signature=[tf.TensorSpec([None], tf.int32)])
def reduce_sum_1d(self, values):
return tf.math.reduce_sum(values)
# reduce_sum_2d (partially dynamic input size, static output size)
# e.g. [[1, 2, 3], [10, 20, 30]] -> [11, 22, 33]
@tf.function(input_signature=[tf.TensorSpec([None, 3], tf.int32)])
def reduce_sum_2d(self, values):
return tf.math.reduce_sum(values, 0)
# add_one (dynamic input size, dynamic output size)
# e.g. [1, 2, 3] -> [2, 3, 4]
@tf.function(input_signature=[tf.TensorSpec([None], tf.int32)])
def add_one(self, values):
return tf.math.add(values, tf.constant(1, dtype=tf.int32))
%%capture
!python -m pip install iree-compiler iree-tools-tf -f https://github.com/google/iree/releases
#@title Import the TensorFlow program into IREE as MLIR
from IPython.display import clear_output
from iree.compiler import tf as tfc
compiler_module = tfc.compile_module(
DynamicShapesModule(), import_only=True,
output_mlir_debuginfo=False)
clear_output() # Skip over TensorFlow's output.
# Print the imported MLIR to see how the compiler views this program.
print("Dynamic Shapes MLIR:\n```\n%s```\n" % compiler_module.decode("utf-8"))
# Save the imported MLIR to disk.
imported_mlir_path = os.path.join(ARTIFACTS_DIR, "dynamic_shapes.mlir")
with open(imported_mlir_path, "wt") as output_file:
output_file.write(compiler_module.decode("utf-8"))
print(f"Wrote MLIR to path '{imported_mlir_path}'")
###Output
Dynamic Shapes MLIR:
```
"builtin.module"() ({
"func.func"() ({
^bb0(%arg0: !iree_input.buffer_view):
%0 = "iree_input.cast.buffer_view_to_tensor"(%arg0) : (!iree_input.buffer_view) -> tensor<?xi32>
%1 = "func.call"(%0) {callee = @__inference_add_one_70} : (tensor<?xi32>) -> tensor<?xi32>
%2 = "iree_input.cast.tensor_to_buffer_view"(%1) : (tensor<?xi32>) -> !iree_input.buffer_view
"func.return"(%2) : (!iree_input.buffer_view) -> ()
}) {function_type = (!iree_input.buffer_view) -> !iree_input.buffer_view, iree.abi = "{\22a\22:[[\22ndarray\22,\22i32\22,1,null]],\22r\22:[[\22ndarray\22,\22i32\22,1,null]],\22v\22:1}", sym_name = "add_one"} : () -> ()
"func.func"() ({
^bb0(%arg0: tensor<?xi32>):
%0 = "mhlo.constant"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
%1 = "chlo.broadcast_add"(%arg0, %0) {broadcast_dimensions = dense<> : tensor<0xi64>} : (tensor<?xi32>, tensor<i32>) -> tensor<?xi32>
"func.return"(%1) : (tensor<?xi32>) -> ()
}) {arg_attrs = [{tf._user_specified_name = "values"}], function_type = (tensor<?xi32>) -> tensor<?xi32>, sym_name = "__inference_add_one_70", sym_visibility = "private", tf._construction_context = "kEagerRuntime", tf._input_shapes = [#tf_type.shape<?>]} : () -> ()
"func.func"() ({
^bb0(%arg0: !iree_input.buffer_view):
%0 = "mhlo.constant"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
%1 = "iree_input.cast.buffer_view_to_tensor"(%arg0) : (!iree_input.buffer_view) -> tensor<?xi32>
%2 = "mhlo.reduce"(%1, %0) ({
^bb0(%arg1: tensor<i32>, %arg2: tensor<i32>):
%4 = "mhlo.add"(%arg1, %arg2) : (tensor<i32>, tensor<i32>) -> tensor<i32>
"mhlo.return"(%4) : (tensor<i32>) -> ()
}) {dimensions = dense<0> : tensor<1xi64>} : (tensor<?xi32>, tensor<i32>) -> tensor<i32>
%3 = "iree_input.cast.tensor_to_buffer_view"(%2) : (tensor<i32>) -> !iree_input.buffer_view
"func.return"(%3) : (!iree_input.buffer_view) -> ()
}) {function_type = (!iree_input.buffer_view) -> !iree_input.buffer_view, iree.abi = "{\22a\22:[[\22ndarray\22,\22i32\22,1,null]],\22r\22:[[\22ndarray\22,\22i32\22,0]],\22v\22:1}", sym_name = "reduce_sum_1d"} : () -> ()
"func.func"() ({
^bb0(%arg0: !iree_input.buffer_view):
%0 = "mhlo.constant"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
%1 = "iree_input.cast.buffer_view_to_tensor"(%arg0) : (!iree_input.buffer_view) -> tensor<?x3xi32>
%2 = "mhlo.reduce"(%1, %0) ({
^bb0(%arg1: tensor<i32>, %arg2: tensor<i32>):
%4 = "mhlo.add"(%arg1, %arg2) : (tensor<i32>, tensor<i32>) -> tensor<i32>
"mhlo.return"(%4) : (tensor<i32>) -> ()
}) {dimensions = dense<0> : tensor<1xi64>} : (tensor<?x3xi32>, tensor<i32>) -> tensor<3xi32>
%3 = "iree_input.cast.tensor_to_buffer_view"(%2) : (tensor<3xi32>) -> !iree_input.buffer_view
"func.return"(%3) : (!iree_input.buffer_view) -> ()
}) {function_type = (!iree_input.buffer_view) -> !iree_input.buffer_view, iree.abi = "{\22a\22:[[\22ndarray\22,\22i32\22,2,null,3]],\22r\22:[[\22ndarray\22,\22i32\22,1,3]],\22v\22:1}", sym_name = "reduce_sum_2d"} : () -> ()
}) : () -> ()
```
Wrote MLIR to path '/tmp/iree/colab_artifacts/dynamic_shapes.mlir'
###Markdown
Test the imported program_Note: you can stop after each step and use intermediate outputs with other tools outside of Colab.__See the [README](https://github.com/google/iree/tree/main/iree/samples/dynamic_shapesinstructions) for more details and example command line instructions._* _The "imported MLIR" can be used by IREE's generic compiler tools_* _The "flatbuffer blob" can be saved and used by runtime applications__The specific point at which you switch from Python to native tools will depend on your project._
###Code
%%capture
!python -m pip install iree-compiler -f https://github.com/google/iree/releases
#@title Compile the imported MLIR further into an IREE VM bytecode module
from iree.compiler import compile_str
# Note: we'll use the cpu (LLVM) backend since it has the best support
# for dynamic shapes among our compiler targets.
flatbuffer_blob = compile_str(compiler_module, target_backends=["cpu"], input_type="mhlo")
# Save the compiled program to disk.
flatbuffer_path = os.path.join(ARTIFACTS_DIR, "dynamic_shapes_cpu.vmfb")
with open(flatbuffer_path, "wb") as output_file:
output_file.write(flatbuffer_blob)
print(f"Wrote compiled program to path '{flatbuffer_path}'")
%%capture
!python -m pip install iree-runtime -f https://github.com/google/iree/releases
#@title Test running the compiled VM module using IREE's runtime
from iree import runtime as ireert
vm_module = ireert.VmModule.from_flatbuffer(flatbuffer_blob)
config = ireert.Config("local-task")
ctx = ireert.SystemContext(config=config)
ctx.add_vm_module(vm_module)
import numpy as np
# Our @tf.functions are accessible by name on the module named 'module'
dynamic_shapes_program = ctx.modules.module
print(dynamic_shapes_program.reduce_sum_1d(np.array([1, 10, 100], dtype=np.int32)).to_host())
print(dynamic_shapes_program.reduce_sum_2d(np.array([[1, 2, 3], [10, 20, 30]], dtype=np.int32)).to_host())
print(dynamic_shapes_program.reduce_sum_2d(np.array([[1, 2, 3], [10, 20, 30], [100, 200, 300]], dtype=np.int32)).to_host())
print(dynamic_shapes_program.add_one(np.array([1, 10, 100], dtype=np.int32)).to_host())
###Output
111
[11 22 33]
[111 222 333]
[ 2 11 101]
###Markdown
Download compilation artifacts
###Code
ARTIFACTS_ZIP = "/tmp/dynamic_shapes_colab_artifacts.zip"
print(f"Zipping '{ARTIFACTS_DIR}' to '{ARTIFACTS_ZIP}' for download...")
!cd {ARTIFACTS_DIR} && zip -r {ARTIFACTS_ZIP} .
# Note: you can also download files using Colab's file explorer
try:
from google.colab import files
print("Downloading the artifacts zip file...")
files.download(ARTIFACTS_ZIP)
except ImportError:
print("Missing google_colab Python package, can't download files")
###Output
Zipping '/tmp/iree/colab_artifacts' to '/tmp/dynamic_shapes_colab_artifacts.zip' for download...
updating: dynamic_shapes.mlir (deflated 80%)
adding: dynamic_shapes_cpu.vmfb (deflated 63%)
Downloading the artifacts zip file...
|
Keras/Diabetes Keras.ipynb | ###Markdown
Iris dataset in Keras library Imports
###Code
import pandas as pd
import numpy as np
import tensorflow as tf
from tensorflow.keras.utils import to_categorical
from sklearn.utils import shuffle
from sklearn.preprocessing import LabelEncoder
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
print("TensorFlow version:", tf.__version__)
###Output
TensorFlow version: 2.2.0
###Markdown
Load data
###Code
# Load data
np_data = pd.read_csv("../data/diabetes.csv").values
# Split data into X and y
X_raw = np_data[:,0:-1].astype(float)
y_raw = np_data[:,-1]
# Shuffle data
X_raw, y_raw = shuffle(X_raw, y_raw, random_state=0)
# Convert class label strings to integers
encoder = LabelEncoder()
encoder.fit(y_raw)
y = encoder.transform(y_raw)
# Normalize data to avoid high input values
scaler = StandardScaler()
scaler.fit(X_raw)
X = scaler.transform(X_raw)
# Convert labels to one-hot vector
y = to_categorical(y, 2)
# Print some stuff
print("Example:")
print(X[0], "->", y_raw[0], "=", y[0])
print("")
print("Data shape:", X.shape)
###Output
Example:
[-0.84488505 2.44447821 0.35643175 1.40909441 -0.69289057 1.38436175
2.784923 -0.95646168] -> YES = [0. 1.]
Data shape: (768, 8)
###Markdown
Train-test split
###Code
# Split data into 80% training and 20% testing
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20)
# Print some stuff
print("Training data shape:", X_train.shape)
print("Testing data shape:", X_test.shape)
###Output
Training data shape: (614, 8)
Testing data shape: (154, 8)
###Markdown
Train and evaluate NN model on all data
###Code
# Create neural network model
model = Sequential()
model.add(Dense(64, input_dim=8, activation="relu", kernel_initializer="he_normal"))
model.add(Dense(2, activation="softmax"))
# Compile model
model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"])
# Train the model on all data
model.fit(X, y, epochs=200, batch_size=32, verbose=0)
# Evaluate on all data
score = model.evaluate(X, y, verbose=0)
# Print results
print("Accuracy: {0:.2f}%".format(score[1]*100))
###Output
Accuracy: 85.29%
###Markdown
Train and evaluate NN model on test data
###Code
# Create neural network model
model = Sequential()
model.add(Dense(64, input_dim=8, activation="relu", kernel_initializer="he_normal"))
model.add(Dense(2, activation="softmax"))
# Compile model
model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"])
# Train the model on training data
model.fit(X_train, y_train, epochs=200, batch_size=32, verbose=0)
# Evaluate on test data
score = model.evaluate(X_test, y_test, verbose=0)
# Print results
print("Accuracy: {0:.2f}%".format(score[1]*100))
###Output
Accuracy: 78.57%
###Markdown
Confusion matrix
###Code
import numpy as np
from sklearn.metrics import confusion_matrix
# Make predictions
y_pred = model.predict(X_test)
# Confusion matrix
conf_mx = confusion_matrix(
np.argmax(y_test,axis=1),
np.argmax(y_pred, axis=1))
print(conf_mx)
###Output
[[89 19]
[14 32]]
###Markdown
Predict new examples
###Code
# Create two new examples
example = [
[6,149,71,34,0,33.6,0.637,48],
[1,83,67,28,0,27.6,0.359,32]
]
# Normalize values
example = scaler.transform(example)
# Make prediction
res = model.predict(example)
print("Prediction:", np.argmax(res, axis=1))
###Output
Prediction: [1 0]
|
Identificando_e_Removendo_Outliers.ipynb | ###Markdown
###Code
%matplotlib inline
import pandas as pd
import matplotlib.pyplot as plt
plt.rc('figure', figsize = (14, 6))
dados = pd.read_csv('aluguel_residencial.csv', sep= ';')
#<img src='Box-Plot.png' width=70%>
dados.boxplot(['Valor'])
dados[dados['Valor'] >= 50000]
valor = dados['Valor']
plt.rc('figure', figsize = (8, 3))
dados = pd.read_csv('aluguel_residencial.csv', sep= ';')
Q1 = valor.quantile(.25)
Q3 = valor.quantile(.75)
IIQ = Q3 - Q1
limite_inferior = Q1 - 1.5 * IIQ
limite_superior = Q3 + 1.5 * IIQ
selecao = (valor >= limite_inferior) & (valor <= limite_superior)
dados_new = dados[selecao]
dados.boxplot(['Valor'])
dados.hist(['Valor'])
dados_new.hist(['Valor'])
# exercicio
data = pd.read_csv('aluguel_amostra.csv', sep = ';')
data.head()
dados.boxplot(['Valor'])
grupo_tipo = dados.groupby('Tipo')
type(grupo_tipo)
grupo_tipo.groups
Q1 = grupo_tipo.quantile(.25)
Q3 = grupo_tipo.quantile(.75)
IIQ = Q3 - Q1
limite_inferior = Q1 - 1.5 * IIQ
limite_superior = Q3 + 1.5 * IIQ
Q1
Q3
limite_inferior
limite_superior
dados_new = pd.DataFrame()
for tipo in grupo_tipo.groups.keys():
eh_tipo = dados['Tipo'] == tipo
eh_dentro_limite = (dados['Valor'] >= limite_inferior[tipo]) and (dados['Valor'] <= limite_superior[tipo])
selecao = eh_tipo & eh_dentro_limite
dados_selecao = dados[selecao]
dados_new = pd.concat([dado_new, dados_selecao])
###Output
_____no_output_____ |
2-resources/BLOG/Python/python_teaching_fall_2018-master/Functions.ipynb | ###Markdown
Functions===A function is a named block of code. You can call it as many times as you want.You can use other people's functions.
###Code
# A simple example.
def greeter():
print("Hello!")
greeter()
###Output
Hello!
###Markdown
Arguments---An argument is a piece of information that you send to a function.
###Code
# Example, with one argument.
def greet_person(name):
print(f"Hello {name}!")
greet_person('philip')
def greet_person(name):
print(f"Hello {name}!")
names = ['eric', 'evan', 'devin', 'philip', 'barack obama', 'abraham', 'psalm']
for name in names:
greet_person(name)
###Output
Hello eric!
Hello evan!
Hello devin!
Hello philip!
Hello barack obama!
Hello abraham!
Hello psalm!
###Markdown
A fun example---
###Code
# Text to speech.
from gtts import gTTS
import os
tts = gTTS(text='Good morning, gandhi.', lang='en')
tts.save("hello.mp3")
os.system("start hello.mp3")
from gtts import gTTS
import os
def say_message(message):
tts = gTTS(text=message, lang='en')
tts.save("message.mp3")
os.system("start message.mp3")
say_message('oink oink!')
names = ['eric', 'evan', 'devin', 'philip', 'barack obama', 'abraham', 'psalm']
for name in names:
say_message(f"Hello, {name}!")
###Output
_____no_output_____
###Markdown
Returning values---You can write a function that does a bunch of work, and then returns something to the line that called the function:
###Code
def get_full_name(f_name, l_name):
"""Return a full name."""
full_name = f"{f_name} {l_name}"
return full_name
full_name = get_full_name('alex', 'honnold')
print(full_name)
###Output
alex honnold
|
loading_data/pandas_features.ipynb | ###Markdown
Pandas can have datetime indexes and work with them
###Code
today = datetime.today()
date_list = [today + timedelta(days=x) for x in range(0, 100)]
df = pd.DataFrame({'values': range(0, 100)}, index=date_list)
df[today:today + timedelta(days=3)]
###Output
_____no_output_____
###Markdown
We can also use partial date objects or another formats. Pandas will try to infer.
###Code
df[str(today.date())]
df[str(today.date())]
###Output
_____no_output_____ |
code_files/Linear Regression on Medical Insurance Dataset.ipynb | ###Markdown
Linear Regression on Medical Insurance Dataset
###Code
import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split
from sklearn.linear_model import LinearRegression
from sklearn.metrics import r2_score
from sklearn.metrics import mean_squared_error
data = pd.read_csv("C:\\Users\\black\\Desktop\\ml_py\\datasets\\insurance.csv")
data.head()
data.info()
###Output
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1338 entries, 0 to 1337
Data columns (total 7 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 age 1338 non-null int64
1 sex 1338 non-null object
2 bmi 1338 non-null float64
3 children 1338 non-null int64
4 smoker 1338 non-null object
5 region 1338 non-null object
6 charges 1338 non-null float64
dtypes: float64(2), int64(2), object(3)
memory usage: 73.3+ KB
###Markdown
Simple Linear Model
###Code
x = data[["age"]].values
y = data.iloc[:,6:].values
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=.2, random_state=40)
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
x_train = sc.fit_transform(x_train)
x_test = sc.transform(x_test)
lr = LinearRegression()
lr.fit(x_train, y_train)
y_pred = lr.predict(x_test)
r2 = r2_score(y_test, y_pred)
r2
from sklearn import metrics
print('Mean Absolute Error:', metrics.mean_absolute_error(y_test, y_pred))
print('Mean Squared Error:', metrics.mean_squared_error(y_test, y_pred))
print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))
###Output
Mean Absolute Error: 9099.67487979512
Mean Squared Error: 137472322.7051458
Root Mean Squared Error: 11724.859176346034
###Markdown
Multi Linear Model
###Code
x = data[["children","bmi"]].values
y = data.iloc[:,6:].values
from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
x_train = sc.fit_transform(x_train)
x_test = sc.transform(x_test)
x_train, x_test, y_train, y_test = train_test_split(x, y, test_size=.2, random_state=40)
lr = LinearRegression()
lr.fit(x_train, y_train)
y_pred = lr.predict(x_test)
r2 = r2_score(y_test, y_pred)
r2
from sklearn import metrics
print('Mean Absolute Error:', metrics.mean_absolute_error(y_test, y_pred))
print('Mean Squared Error:', metrics.mean_squared_error(y_test, y_pred))
print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))
###Output
Mean Absolute Error: 9426.280328467989
Mean Squared Error: 143983516.1469823
Root Mean Squared Error: 11999.31315313432
|
demos/location-based-recommendation/04-nuclio-process_user_location.ipynb | ###Markdown
Nuclio - Process user location signal Setup the environment
###Code
# nuclio: ignore
import nuclio
###Output
_____no_output_____
###Markdown
Define environment variables
###Code
import os
# nuclio: ignore
os.environ['STORES_TABLE'] = os.path.join(os.getenv('V3IO_USERNAME', 'iguazio'), 'stores')
os.environ['CUSTOMERS_TABLE'] = os.path.join(os.getenv('V3IO_USERNAME', 'iguazio'), 'customers')
os.environ['CUSTOMERS_STREAM'] = os.path.join(os.getenv('V3IO_USERNAME', 'iguazio'), 'customers_stream')
os.environ['PREDICTIONS_STREAM'] = os.path.join(os.getenv('V3IO_USERNAME', 'iguazio'), 'predictions')
# DB Acess
%nuclio env V3IO_API=${V3IO_FRAMESD}
%nuclio env V3IO_ACCESS_KEY=${V3IO_ACCESS_KEY}
# Customers
%nuclio env COSTUMERS_STREAM=${CUSTOMERS_STREAM}
%nuclio env COSTUMERS_TABLE=${CUSTOMERS_TABLE}
%nuclio env STORES_TABLE=${STORES_TABLE}
# Predictions
%nuclio env PREDICTIONS_STREAM=${PREDICTIONS_STREAM}
%nuclio env PREDICTION_SERVER=http://prediction-server:8080
###Output
%nuclio: setting 'V3IO_API' environment variable
%nuclio: setting 'V3IO_ACCESS_KEY' environment variable
%nuclio: setting 'COSTUMERS_STREAM' environment variable
%nuclio: setting 'COSTUMERS_TABLE' environment variable
%nuclio: setting 'STORES_TABLE' environment variable
%nuclio: setting 'PREDICTIONS_STREAM' environment variable
%nuclio: setting 'PREDICTION_SERVER' environment variable
###Markdown
Base image
###Code
%nuclio config spec.build.baseImage = "python:3.6-jessie"
###Output
%nuclio: setting spec.build.baseImage to 'python:3.6-jessie'
###Markdown
Set cron trigger to read from stream
###Code
%nuclio config spec.triggers.secs.kind = "cron"
%nuclio config spec.triggers.secs.attributes.interval = "1m"
###Output
%nuclio: setting spec.triggers.secs.kind to 'cron'
%nuclio: setting spec.triggers.secs.attributes.interval to '1m'
###Markdown
Install packages
###Code
%%nuclio cmd -c
pip install v3io_frames
pip install v3io==0.1.1 --upgrade
pip install requests
pip install pandas
###Output
_____no_output_____
###Markdown
Function code
###Code
import json
import os
import requests
import time
# Data handling
import pandas as pd
# DB
import v3io
import v3io.dataplane
import v3io.logger
import v3io_frames as v3f
###Output
_____no_output_____
###Markdown
Init context
###Code
def init_context(context):
# DB Contexts
v3c = v3io.dataplane.Context(v3io.logger.Logger('DEBUG')).new_session().new_container('users')
setattr(context, 'v3c', v3c)
v3c_frames = v3f.Client('framesd:8081', container='users')
setattr(context, 'v3f', v3c_frames)
# DB Tables
customers_table = os.environ['COSTUMERS_TABLE']
setattr(context, 'customers', customers_table)
stores_table = os.environ['STORES_TABLE']
setattr(context, 'stores', stores_table)
predictions_stream = os.environ['PREDICTIONS_STREAM']
setattr(context, 'predictions', predictions_stream)
customers_stream = os.environ['COSTUMERS_STREAM']
setattr(context, 'customers_stream', customers_stream)
# Prediction server
prediction_server = os.getenv('PREDICTION_SERVER')
setattr(context, 'prediction_server', prediction_server)
###Output
_____no_output_____
###Markdown
Helper functions
###Code
def is_customer_in_store(customer, context) -> bool:
store_location = customer['location']
store = context.v3f.read('kv', context.stores, filter=f'__name=="{store_location}"')
return not store.empty
def is_customer_out_of_store(context, new_customer_locations):
if not new_customer_locations.empty:
users = new_customer_locations['id'].values.astype('int').astype('str')
filter_line = str(list(users))
filter_line = f'__name IN ({filter_line[1:-1]})'
old_customer_locations = context.v3f.read('kv', context.customers, columns=['location'], filter=filter_line)
old_customer_locations['is_store'] = old_customer_locations.apply(is_customer_in_store, args=[context], axis=1)
return old_customer_locations[old_customer_locations['is_store'] == True]['location']
def update_customer_location(context, customer_id: str, location: str):
context.v3f.execute('kv', context.customers , 'update', args={'key':customer_id, 'expression': f'SET location="{location}"', 'condition':''})
def update_store_count(customer, context, is_add=True):
operator = '+' if is_add else '-'
context.v3f.execute('kv', context.stores , 'update', args={'key': customer, 'expression': f'SET customers=customers{operator}1', 'condition':''})
def save_predictions(context, customer_id: str, prediction: pd.DataFrame):
context.v3f.write('tsdb', context.predictions, prediction)
###Output
_____no_output_____
###Markdown
Handler
###Code
def handler(context, event):
# Get latest customer locations from the customers stream
customers_stream = context.v3f.read('tsdb', context.customers_stream, start='now-1m')
if not customers_stream.empty:
# Has anyone moved out of any store?
stores_to_update = is_customer_out_of_store(context, customers_stream)
stores_to_update.apply(update_store_count, args=[context, False])
# Update the customer's new location
[update_customer_location(context, str(int(customer['id'])), customer['location']) for idx, customer in customers_stream.iterrows()]
# Get all customers that are in stores
customers_stream['is_store'] = customers_stream.apply(is_customer_in_store, args=[context], axis=1)
customers_stream = customers_stream[customers_stream['is_store']]
# Update customers in stores count
customers_stream['update_stores'] = customers_stream['location'].apply(update_store_count, args=[context])
context.logger.debug(customers_stream)
[requests.post(context.prediction_server, json={'id': str(int(customer['id'])), 'store': str(customer['location'])}) for idx, customer in customers_stream.iterrows()]
# nuclio: ignore
init_context(context)
# nuclio: ignore
event = nuclio.Event()
handler(context, event)
%nuclio deploy -n process_user_location -p recommendation_engine -c
###Output
[nuclio.deploy] 2019-08-11 13:38:34,372 (info) Building processor image
[nuclio.deploy] 2019-08-11 13:39:04,748 (info) Pushing image
[nuclio.deploy] 2019-08-11 13:39:18,908 (info) Build complete
[nuclio.deploy] 2019-08-11 13:39:22,100 done creating process-user-location, function address: 3.120.15.118:32537
%nuclio: function deployed
|
7.3-developing-and-evaluating-chunkers.ipynb | ###Markdown
开发和评估分块器 读取 IOB 格式与 CoNLL2000 分块语料库使用转换函数 [chunk.conllstr2tree()](https://www.nltk.org/_modules/nltk/chunk/util.htmlconllstr2tree) 可以将 IOB 格式的字符串转换成树表示。此外，它允许我们选择使用语料库提供的三种块类型：NP、VP 和 PP 的任何子集。
###Code
import nltk
text = """
he PRP B-NP
accepted VBD B-VP
the DT B-NP
position NN I-NP
of IN B-PP
vice NN B-NP
chairman NN I-NP
of IN B-PP
Carlyle NNP B-NP
Group NNP I-NP
, , O
a DT B-NP
merchant NN I-NP
banking NN I-NP
concern NN I-NP
. . O
"""
nltk.chunk.conllstr2tree(text, chunk_types=['NP']).draw()
###Output
_____no_output_____
###Markdown
 NLTK 的 corpus 模块包含了大量已分块的文本。CoNLL2000 语料库包含 27 万词的《华尔街日报》文本，分为“训练”和“测试”两部分，标注有词性标记和 IOB 格式分块标记：
###Code
from nltk.corpus import conll2000
print(conll2000.chunked_sents('train.txt')[99])
###Output
(S
(PP Over/IN)
(NP a/DT cup/NN)
(PP of/IN)
(NP coffee/NN)
,/,
(NP Mr./NNP Stone/NNP)
(VP told/VBD)
(NP his/PRP$ story/NN)
./.)
###Markdown
正如你看到的，CoNLL2000 语料库包含了三种块类型：NP 块如 a cup；VP 块如 told；PP 块如 of。由于现在我们唯一感兴趣的是 NP 块，我们可以使用 chunk_types 参数选择它：
###Code
test_sent = conll2000.chunked_sents('train.txt', chunk_types=['NP'])[99]
print(type(test_sent))
print(test_sent)
###Output
<class 'nltk.tree.Tree'>
(S
Over/IN
(NP a/DT cup/NN)
of/IN
(NP coffee/NN)
,/,
(NP Mr./NNP Stone/NNP)
told/VBD
(NP his/PRP$ story/NN)
./.)
###Markdown
简单评估和基准首先，我们为不创建任何块的分块器建立一个基准（baseline）：
###Code
cp = nltk.RegexpParser('')
test_sents = conll2000.chunked_sents('test.txt', chunk_types=['NP'])
print(cp.parse(test_sent))
print(cp.evaluate(test_sents))
###Output
(S
Over/IN
(NP a/DT cup/NN)
of/IN
(NP coffee/NN)
,/,
(NP Mr./NNP Stone/NNP)
told/VBD
(NP his/PRP$ story/NN)
./.)
ChunkParse score:
IOB Accuracy: 43.4%%
Precision: 0.0%%
Recall: 0.0%%
F-Measure: 0.0%%
###Markdown
IOB 标记准确性表明超过三分之一的词被标注为 O，即没有出现在 NP 块中。然而，由于我们的标注器没有找到任何块，其精度、召回率和 F 度量均为零。现在让我们尝试一个初级的正则表达式分类器，查找以名词短语标记的特征字母（如 CD、DT 和 JJ）开头的标记：
###Code
grammar = r'NP: {<[CDJNP].*>+}'
cp = nltk.RegexpParser(grammar)
print(cp.parse(test_sent))
print(cp.evaluate(test_sents))
###Output
(S
Over/IN
(NP (NP a/DT cup/NN))
of/IN
(NP (NP coffee/NN))
,/,
(NP (NP Mr./NNP Stone/NNP))
told/VBD
(NP (NP his/PRP$ story/NN))
./.)
ChunkParse score:
IOB Accuracy: 87.7%%
Precision: 70.6%%
Recall: 67.8%%
F-Measure: 69.2%%
###Markdown
这种方法达到了不错的结果，但是我们可以采用更多数据驱动的方法改善它。这里我们定义了 UnigramChunker 类，使用 unigram 标注器给句子加块标记。这个类实现了 [nltk.ChunkParserI](https://www.nltk.org/_modules/nltk/chunk/api.htmlChunkParserI) 接口，定义了两个方法：一个构造函数，当我们建立新的 UnigramChunker 时调用；一个 parse 方法，用来给新句子分块。构造函数需要训练句子的一个链表，每个句子都是块树的形式。它首先通过 tree2conlltags 方法将块树转换成 IOB 标记，然后训练一个基于词性标记的 unigram 块标注器。parse 方法取一个已标注的句子作为输入，首先提取词性标记，然后使用在构造函数中训练过的标注器为词性标记标注 IOB 标记。接下来将块标记与原句组合，产生 conlltags。最后使用 conlltags2tree 将结果转换成一个块树。
###Code
class UnigramChunker(nltk.ChunkParserI):
def __init__(self, train_sents):
train_data = [[(t, c) for w, t, c in nltk.chunk.tree2conlltags(sent)]
for sent in train_sents]
self.tagger = nltk.UnigramTagger(train_data)
def parse(self, sentence):
pos_tags = [pos for (_, pos) in sentence]
tagged_pos_tags = self.tagger.tag(pos_tags)
chunktags = [chunktag for (pos, chunktag) in tagged_pos_tags]
conlltags = [(word, pos, chunktag) for ((word, pos), chunktag)
in zip(sentence, chunktags)]
return nltk.chunk.conlltags2tree(conlltags)
train_sents = conll2000.chunked_sents('train.txt', chunk_types=['NP'])
test_sents = conll2000.chunked_sents('test.txt', chunk_types=['NP'])
unigram_chunker = UnigramChunker(train_sents)
print(unigram_chunker.evaluate(test_sents))
###Output
ChunkParse score:
IOB Accuracy: 92.9%%
Precision: 79.9%%
Recall: 86.8%%
F-Measure: 83.2%%
###Markdown
这个分块器相当不错，达到整体 F 度量 83% 的得分。现在我们来分析一下 unigram 标注器给每个词性标记分配了什么样的块标记：
###Code
postags = sorted(set(pos for sent in train_sents
for (_, pos) in sent.leaves()))
print(unigram_chunker.tagger.tag(postags))
###Output
[('#', 'B-NP'), ('$', 'B-NP'), ("''", 'O'), ('(', 'O'), (')', 'O'), (',', 'O'), ('.', 'O'), (':', 'O'), ('CC', 'O'), ('CD', 'I-NP'), ('DT', 'B-NP'), ('EX', 'B-NP'), ('FW', 'I-NP'), ('IN', 'O'), ('JJ', 'I-NP'), ('JJR', 'B-NP'), ('JJS', 'I-NP'), ('MD', 'O'), ('NN', 'I-NP'), ('NNP', 'I-NP'), ('NNPS', 'I-NP'), ('NNS', 'I-NP'), ('PDT', 'B-NP'), ('POS', 'B-NP'), ('PRP', 'B-NP'), ('PRP$', 'B-NP'), ('RB', 'O'), ('RBR', 'O'), ('RBS', 'B-NP'), ('RP', 'O'), ('SYM', 'O'), ('TO', 'O'), ('UH', 'O'), ('VB', 'O'), ('VBD', 'O'), ('VBG', 'O'), ('VBN', 'O'), ('VBP', 'O'), ('VBZ', 'O'), ('WDT', 'B-NP'), ('WP', 'B-NP'), ('WP$', 'B-NP'), ('WRB', 'O'), ('``', 'O')]
###Markdown
可以发现大多数标点符号都出现在 NP 块外，除了货币符号 `` 和 `$`；限定词（DT）和所有格（PRP`$` 和 WP`$`）出现在 NP 块的开头，而名词类型（NN，NNP，NNPS，NNS）大多出现在 NP 块内。我们对 unigram 分块器稍作修改，建立一个 bigram 分块器，性能略有提升：
###Code
class BigramChunker(nltk.ChunkParserI):
def __init__(self, train_sents):
train_data = [[(t, c) for w, t, c in nltk.chunk.tree2conlltags(sent)]
for sent in train_sents]
self.tagger = nltk.BigramTagger(train_data)
def parse(self, sentence):
pos_tags = [pos for (_, pos) in sentence]
tagged_pos_tags = self.tagger.tag(pos_tags)
chunktags = [chunktag for (pos, chunktag) in tagged_pos_tags]
conlltags = [(word, pos, chunktag) for ((word, pos), chunktag)
in zip(sentence, chunktags)]
return nltk.chunk.conlltags2tree(conlltags)
train_sents = conll2000.chunked_sents('train.txt', chunk_types=['NP'])
test_sents = conll2000.chunked_sents('test.txt', chunk_types=['NP'])
bigram_chunker = BigramChunker(train_sents)
print(bigram_chunker.evaluate(test_sents))
###Output
ChunkParse score:
IOB Accuracy: 93.3%%
Precision: 82.3%%
Recall: 86.8%%
F-Measure: 84.5%%
###Markdown
训练基于分类器的分块器无论是基于正则表达式的分块器还是 n-gram 分块器，决定创建什么块完全基于词性标记。然而有时词性标记不足以确定一个句子应如何分块。例如：a. Joey/NN sold/VBD the/DT farmer/NN rice/NN ./.b. Nick/NN broke/VBD my/DT computer/NN monitor/NN ./.这两句话的词性标记相同，但分块方式不同。第一句中 the farmer 和 rice 都是单独的块，而第二个句子中相应的部分 the computer monitor 是一个单独的块。因此，为了最大限度地提升分块的性能，我们需要使用词的内容信息作为词性标注的补充。我们包含词的内容信息的方法之一是使用基于**分类器**的标注器对句子分块。在下面的例子中包括两个类：第一个类与 6.1 节中的 ConsecutivePosTagger 类似，仅有的区别在于使用 MaxentClassifier 代替 NaiveBayesClassifier；第二个类是标注器类的一个包装器，将它变成一个分块器。
###Code
class ConsecutiveNPChunkTagger(nltk.TaggerI):
def __init__(self, train_sents):
train_set = []
for tagged_sent in train_sents:
untagged_sent = nltk.tag.untag(tagged_sent)
history = []
for i, (word, tag) in enumerate(tagged_sent):
featureset = npchunk_features(untagged_sent, i, history)
train_set.append((featureset, tag))
history.append(tag)
self.classifier = nltk.MaxentClassifier.train(train_set, trace=0)
def tag(self, sentence):
history = []
for i, word in enumerate(sentence):
featureset = npchunk_features(sentence, i, history)
tag = self.classifier.classify(featureset)
history.append(tag)
return zip(sentence, history)
class ConsecutiveNPChunker(nltk.ChunkParserI):
def __init__(self, train_sents):
tagged_sents = [[((w, t), c) for (w, t, c) in
nltk.chunk.tree2conlltags(sent)]
for sent in train_sents]
self.tagger = ConsecutiveNPChunkTagger(tagged_sents)
def parse(self, sentence):
tagged_sents = self.tagger.tag(sentence)
conlltags = [(w, t, c) for ((w, t), c) in tagged_sents]
return nltk.chunk.conlltags2tree(conlltags)
###Output
_____no_output_____
###Markdown
还需要定义用到的特征提取器 npchunk_features。首先，我们定义一个简单的特征提取器，它只提供当前标识符的词性标记。利用这个体征提取器的分块器性能与 unigram 分块器非常类似：
###Code
def npchunk_features(sentence, i, history):
word, pos = sentence[i]
return {'pos': pos}
chunker = ConsecutiveNPChunker(train_sents)
print(chunker.evaluate(test_sents))
###Output
ChunkParse score:
IOB Accuracy: 92.9%%
Precision: 79.9%%
Recall: 86.8%%
F-Measure: 83.2%%
###Markdown
接着我们再添加一个特征：前面词的词性标记。添加此特征允许分类器模拟相邻标记间的相互作用，由此产生的分块器与 bigram 分块器非常接近：
###Code
def npchunk_features(sentence, i, history):
word, pos = sentence[i]
if i == 0:
prevword, prevpos = '<START>', '<START>'
else:
prevword, prevpos = sentence[i - 1]
return {'pos': pos, 'prevpos': prevpos}
chunker = ConsecutiveNPChunker(train_sents)
print(chunker.evaluate(test_sents))
###Output
ChunkParse score:
IOB Accuracy: 93.6%%
Precision: 82.0%%
Recall: 87.2%%
F-Measure: 84.6%%
###Markdown
下一步，我们尝试把当前词增加为特征，可以发现这个特征确实提高了分块器的性能，大约 1.5 个百分点：
###Code
def npchunk_features(sentence, i, history):
word, pos = sentence[i]
if i == 0:
prevword, prevpos = '<START>', '<START>'
else:
prevword, prevpos = sentence[i - 1]
return {'pos': pos, 'word': word, 'prevpos': prevpos}
chunker = ConsecutiveNPChunker(train_sents)
print(chunker.evaluate(test_sents))
###Output
ChunkParse score:
IOB Accuracy: 94.6%%
Precision: 84.6%%
Recall: 89.8%%
F-Measure: 87.1%%
###Markdown
最后，我们尝试用多种附加特征扩展特征提取器，例如：预取特征、配对功能和复杂的语境特征等：
###Code
def npchunk_features(sentence, i, history):
word, pos = sentence[i]
if i == 0:
prevword, prevpos = '<START>', '<START>'
else:
prevword, prevpos = sentence[i - 1]
if i == len(sentence) - 1:
nextword, nextpos = '<END', 'END>'
else:
nextword, nextpos = sentence[i + 1]
return {'pos': pos,
'word': word,
'prevpos': prevpos,
'nextpos': nextpos,
'prevpos+pos': '%s+%s' % (prevpos, pos),
'pos+nextpos': '%s+%s' % (pos, nextpos),
'tags-since-dt': tags_since_dt(sentence, i)}
def tags_since_dt(sentence, i):
tags = set()
for word, pos in sentence[:i]:
if pos == 'DT':
tags = set()
else:
tags.add(pos)
return '+'.join(sorted(tags))
chunker = ConsecutiveNPChunker(train_sents)
print(chunker.evaluate(test_sents))
###Output
ChunkParse score:
IOB Accuracy: 96.0%%
Precision: 88.3%%
Recall: 91.1%%
F-Measure: 89.7%%
|
Code/nn.ipynb | ###Markdown
Basic Instructions1. Enter your Name and UID in the provided space.2. Do the assignment in the notebook itself3. you are free to use Google Colab Name: **Arpit Aggarwal** UID: **116747189** In the first part, you will implement all the functions required to build a two layer neural network.In the next part, you will use these functions for image and text classification. Provide your code at the appropriate placeholders. 1. Packages
###Code
import numpy as np
import matplotlib.pyplot as plt
import h5py
import scipy
from PIL import Image
from scipy import ndimage
###Output
_____no_output_____
###Markdown
2. Layer Initialization **Exercise:** Create and initialize the parameters of the 2-layer neural network. Use random initialization for the weight matrices and zero initialization for the biases.
###Code
def initialize_parameters(n_x, n_h, n_y):
"""
Argument:
n_x -- size of the input layer
n_h -- size of the hidden layer
n_y -- size of the output layer
Returns:
parameters -- python dictionary containing your parameters:
W1 -- weight matrix of shape (n_h, n_x)
b1 -- bias vector of shape (n_h, 1)
W2 -- weight matrix of shape (n_y, n_h)
b2 -- bias vector of shape (n_y, 1)
"""
np.random.seed(1)
### START CODE HERE ### (≈ 4 lines of code)
W1 = np.random.randn(n_h, n_x) * 0.01
b1 = np.zeros(shape=(n_h, 1))
W2 = np.random.randn(n_y, n_h) * 0.01
b2 = np.zeros(shape=(n_y, 1))
### END CODE HERE ###
assert(W1.shape == (n_h, n_x))
assert(b1.shape == (n_h, 1))
assert(W2.shape == (n_y, n_h))
assert(b2.shape == (n_y, 1))
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2}
return parameters
parameters = initialize_parameters(3,2,1)
print("W1 = " + str(parameters["W1"]))
print("b1 = " + str(parameters["b1"]))
print("W2 = " + str(parameters["W2"]))
print("b2 = " + str(parameters["b2"]))
###Output
W1 = [[ 0.01624345 -0.00611756 -0.00528172]
[-0.01072969 0.00865408 -0.02301539]]
b1 = [[0.]
[0.]]
W2 = [[ 0.01744812 -0.00761207]]
b2 = [[0.]]
###Markdown
**Expected output**: **W1** [[ 0.01624345 -0.00611756 -0.00528172] [-0.01072969 0.00865408 -0.02301539]] **b1** [[ 0.] [ 0.]] **W2** [[ 0.01744812 -0.00761207]] **b2** [[ 0.]] 3. Forward Propagation Now that you have initialized your parameters, you will do the forward propagation module. You will start by implementing some basic functions that you will use later when implementing the model. You will complete three functions in this order:- LINEAR- LINEAR -> ACTIVATION where ACTIVATION will be either ReLU or Sigmoid.The linear module computes the following equation:$$Z = WA+b\tag{4}$$ 3.1 Exercise: Build the linear part of forward propagation.
###Code
def linear_forward(A, W, b):
"""
Implement the linear part of a layer's forward propagation.
Arguments:
A -- activations from previous layer (or input data): (size of previous layer, number of examples)
W -- weights matrix: numpy array of shape (size of current layer, size of previous layer)
b -- bias vector, numpy array of shape (size of the current layer, 1)
Returns:
Z -- the input of the activation function, also called pre-activation parameter
cache -- a python dictionary containing "A", "W" and "b" ; stored for computing the backward pass efficiently
"""
### START CODE HERE ### (≈ 1 line of code)
Z = np.dot(W, A) + b
### END CODE HERE ###
assert(Z.shape == (W.shape[0], A.shape[1]))
cache = (A, W, b)
return Z, cache
np.random.seed(1)
A = np.random.randn(3,2)
W = np.random.randn(1,3)
b = np.random.randn(1,1)
Z, linear_cache = linear_forward(A, W, b)
print("Z = " + str(Z))
###Output
Z = [[ 3.26295337 -1.23429987]]
###Markdown
**Expected output**: **Z** [[ 3.26295337 -1.23429987]] 3.2 - Linear-Activation ForwardIn this notebook, you will use two activation functions:- **Sigmoid**: $\sigma(Z) = \sigma(W A + b) = \frac{1}{ 1 + e^{-(W A + b)}}$. Write the code for the `sigmoid` function. This function returns **two** items: the activation value "`a`" and a "`cache`" that contains "`Z`" (it's what we will feed in to the corresponding backward function). To use it you could just call: ``` pythonA, activation_cache = sigmoid(Z)```- **ReLU**: The mathematical formula for ReLu is $A = RELU(Z) = max(0, Z)$. Write the code for the `relu` function. This function returns **two** items: the activation value "`A`" and a "`cache`" that contains "`Z`" (it's what we will feed in to the corresponding backward function). To use it you could just call:``` pythonA, activation_cache = relu(Z)**Exercise**: - Implement the activation functions- Build the linear activation part of forward propagation. Mathematical relation is: $A = g(Z) = g(WA_{prev} +b)$
###Code
def sigmoid(Z):
"""
Implements the sigmoid activation in numpy
Arguments:
Z -- numpy array of any shape
Returns:
A -- output of sigmoid(z), same shape as Z
cache -- returns Z, useful during backpropagation
"""
### START CODE HERE ### (≈ 2 line of code)
A = 1.0 / (1.0 + np.exp(-Z))
cache = Z
### END CODE HERE ###
return A, cache
def relu(Z):
"""
Implement the RELU function.
Arguments:
Z -- Output of the linear layer, of any shape
Returns:
A -- Post-activation parameter, of the same shape as Z
cache -- returns Z, useful during backpropagation
"""
### START CODE HERE ### (≈ 2 line of code)
A = np.maximum(0, Z)
cache = Z
### END CODE HERE ###
assert(A.shape == Z.shape)
return A, cache
def linear_activation_forward(A_prev, W, b, activation):
"""
Implement the forward propagation for the LINEAR->ACTIVATION layer
Arguments:
A_prev -- activations from previous layer (or input data): (size of previous layer, number of examples)
W -- weights matrix: numpy array of shape (size of current layer, size of previous layer)
b -- bias vector, numpy array of shape (size of the current layer, 1)
activation -- the activation to be used in this layer, stored as a text string: "sigmoid" or "relu"
Returns:
A -- the output of the activation function, also called the post-activation value
cache -- a python dictionary containing "linear_cache" and "activation_cache";
stored for computing the backward pass efficiently
"""
if activation == "sigmoid":
# Inputs: "A_prev, W, b". Outputs: "A, activation_cache".
### START CODE HERE ### (≈ 2 lines of code)
Z, linear_cache = linear_forward(A_prev, W, b)
A, activation_cache = sigmoid(Z)
### END CODE HERE ###
elif activation == "relu":
# Inputs: "A_prev, W, b". Outputs: "A, activation_cache".
### START CODE HERE ### (≈ 2 lines of code)
Z, linear_cache = linear_forward(A_prev, W, b)
A, activation_cache = relu(Z)
### END CODE HERE ###
assert (A.shape == (W.shape[0], A_prev.shape[1]))
cache = (linear_cache, activation_cache)
return A, cache
np.random.seed(2)
A_prev = np.random.randn(3,2)
W = np.random.randn(1,3)
b = np.random.randn(1,1)
A, linear_activation_cache = linear_activation_forward(A_prev, W, b, activation = "sigmoid")
print("With sigmoid: A = " + str(A))
A, linear_activation_cache = linear_activation_forward(A_prev, W, b, activation = "relu")
print("With ReLU: A = " + str(A))
###Output
With sigmoid: A = [[0.96890023 0.11013289]]
With ReLU: A = [[3.43896131 0. ]]
###Markdown
**Expected output**: **With sigmoid: A ** [[ 0.96890023 0.11013289]] **With ReLU: A ** [[ 3.43896131 0. ]] 4 - Loss functionNow you will implement forward and backward propagation. You need to compute the loss, because you want to check if your model is actually learning.**Exercise**: Compute the cross-entropy loss $J$, using the following formula: $$-\frac{1}{m} \sum\limits_{i = 1}^{m} (y^{(i)}\log\left(a^{ (i)}\right) + (1-y^{(i)})\log\left(1- a^{(i)}\right)) \tag{7}$$
###Code
# GRADED FUNCTION: compute_loss
def compute_loss(A, Y):
"""
Implement the loss function defined by equation (7).
Arguments:
A -- probability vector corresponding to your label predictions, shape (1, number of examples)
Y -- true "label" vector (for example: containing 0 if non-cat, 1 if cat), shape (1, number of examples)
Returns:
loss -- cross-entropy loss
"""
m = Y.shape[1]
# Compute loss from aL and y.
### START CODE HERE ### (≈ 1 lines of code)
loss = (-1.0 / m) * np.sum((Y * np.log(A)) + ((1.0 - Y) * np.log(1.0 - A)))
### END CODE HERE ###
loss = np.squeeze(loss) # To make sure your loss's shape is what we expect (e.g. this turns [[17]] into 17).
assert(loss.shape == ())
return loss
Y = np.asarray([[1, 1, 1]])
A = np.array([[.8,.9,0.4]])
print("loss = " + str(compute_loss(A, Y)))
###Output
loss = 0.41493159961539694
###Markdown
**Expected Output**: **loss** 0.41493159961539694 5 - Backward propagation moduleJust like with forward propagation, you will implement helper functions for backpropagation. Remember that back propagation is used to calculate the gradient of the loss function with respect to the parameters. Now, similar to forward propagation, you are going to build the backward propagation in two steps:- LINEAR backward- LINEAR -> ACTIVATION backward where ACTIVATION computes the derivative of either the ReLU or sigmoid activation 5.1 - Linear backward
###Code
# GRADED FUNCTION: linear_backward
def linear_backward(dZ, cache):
"""
Implement the linear portion of backward propagation for a single layer (layer l)
Arguments:
dZ -- Gradient of the loss with respect to the linear output (of current layer l)
cache -- tuple of values (A_prev, W, b) coming from the forward propagation in the current layer
Returns:
dA_prev -- Gradient of the loss with respect to the activation (of the previous layer l-1), same shape as A_prev
dW -- Gradient of the loss with respect to W (current layer l), same shape as W
db -- Gradient of the loss with respect to b (current layer l), same shape as b
"""
A_prev, W, b = cache
m = A_prev.shape[1]
### START CODE HERE ### (≈ 3 lines of code)
dA_prev = np.dot(W.T, dZ)
dW = np.dot(dZ, A_prev.T)
db = np.array([np.sum(dZ, axis = 1)]).T
### END CODE HERE ###
assert (dA_prev.shape == A_prev.shape)
assert (dW.shape == W.shape)
assert (db.shape == b.shape)
return dA_prev, dW, db
np.random.seed(1)
dZ = np.random.randn(1,2)
A = np.random.randn(3,2)
W = np.random.randn(1,3)
b = np.random.randn(1,1)
linear_cache = (A, W, b)
dA_prev, dW, db = linear_backward(dZ, linear_cache)
print ("dA_prev = "+ str(dA_prev))
print ("dW = " + str(dW))
print ("db = " + str(db))
###Output
dA_prev = [[ 0.51822968 -0.19517421]
[-0.40506361 0.15255393]
[ 2.37496825 -0.89445391]]
dW = [[-0.2015379 2.81370193 3.2998501 ]]
db = [[1.01258895]]
###Markdown
**Expected Output**: **dA_prev** [[ 0.51822968 -0.19517421] [-0.40506361 0.15255393] [ 2.37496825 -0.89445391]] **dW** [[-0.2015379 2.81370193 3.2998501 ]] **db** [[1.01258895]] 5.2 - Linear Activation backwardNext, you will create a function that merges the two helper functions: **`linear_backward`** and the backward step for the activation **`linear_activation_backward`**. Before implementing `linear_activation_backward`, you need to implement two backward functions for each activations:- **`sigmoid_backward`**: Implements the backward propagation for SIGMOID unit. You can call it as follows:```pythondZ = sigmoid_backward(dA, activation_cache)```- **`relu_backward`**: Implements the backward propagation for RELU unit. You can call it as follows:```pythondZ = relu_backward(dA, activation_cache)```If $g(.)$ is the activation function, `sigmoid_backward` and `relu_backward` compute $$dZ^{[l]} = dA^{[l]} * g'(Z^{[l]}) \tag{11}$$. **Exercise**: - Implement the backward functions for the relu and sigmoid activation layer.- Implement the backpropagation for the *LINEAR->ACTIVATION* layer.
###Code
def relu_backward(dA, cache):
"""
Implement the backward propagation for a single RELU unit.
Arguments:
dA -- post-activation gradient, of any shape
cache -- 'Z' where we store for computing backward propagation efficiently
Returns:
dZ -- Gradient of the loss with respect to Z
"""
Z = cache
dZ = np.array(dA, copy=True) # just converting dz to a correct object.
### START CODE HERE ### (≈ 1 line of code)
dZ = dA * np.where(Z <= 0, 0, 1)
### END CODE HERE ###
assert (dZ.shape == Z.shape)
return dZ
def sigmoid_backward(dA, cache):
"""
Implement the backward propagation for a single SIGMOID unit.
Arguments:
dA -- post-activation gradient, of any shape
cache -- 'Z' where we store for computing backward propagation efficiently
Returns:
dZ -- Gradient of the loss with respect to Z
"""
Z = cache
### START CODE HERE ### (≈ 2 line of code)
sigmoid_derivative = sigmoid(Z)[0] * (1.0 - sigmoid(Z)[0])
dZ = dA * sigmoid_derivative
### END CODE HERE ###
assert (dZ.shape == Z.shape)
return dZ
# GRADED FUNCTION: linear_activation_backward
def linear_activation_backward(dA, cache, activation):
"""
Implement the backward propagation for the LINEAR->ACTIVATION layer.
Arguments:
dA -- post-activation gradient for current layer l
cache -- tuple of values (linear_cache, activation_cache) we store for computing backward propagation efficiently
activation -- the activation to be used in this layer, stored as a text string: "sigmoid" or "relu"
Returns:
dA_prev -- Gradient of the loss with respect to the activation (of the previous layer l-1), same shape as A_prev
dW -- Gradient of the loss with respect to W (current layer l), same shape as W
db -- Gradient of the loss with respect to b (current layer l), same shape as b
"""
linear_cache, activation_cache = cache
if activation == "relu":
### START CODE HERE ### (≈ 2 lines of code)
dZ = relu_backward(dA, activation_cache)
dA_prev, dW, db = linear_backward(dZ, linear_cache)
### END CODE HERE ###
elif activation == "sigmoid":
### START CODE HERE ### (≈ 2 lines of code)
dZ = sigmoid_backward(dA, activation_cache)
dA_prev, dW, db = linear_backward(dZ, linear_cache)
### END CODE HERE ###
return dA_prev, dW, db
np.random.seed(2)
dA = np.random.randn(1,2)
A = np.random.randn(3,2)
W = np.random.randn(1,3)
b = np.random.randn(1,1)
Z = np.random.randn(1,2)
linear_cache = (A, W, b)
activation_cache = Z
linear_activation_cache = (linear_cache, activation_cache)
dA_prev, dW, db = linear_activation_backward(dA, linear_activation_cache, activation = "sigmoid")
print ("sigmoid:")
print ("dA_prev = "+ str(dA_prev))
print ("dW = " + str(dW))
print ("db = " + str(db) + "\n")
dA_prev, dW, db = linear_activation_backward(dA, linear_activation_cache, activation = "relu")
print ("relu:")
print ("dA_prev = "+ str(dA_prev))
print ("dW = " + str(dW))
print ("db = " + str(db))
###Output
sigmoid:
dA_prev = [[ 0.11017994 0.01105339]
[ 0.09466817 0.00949723]
[-0.05743092 -0.00576154]]
dW = [[ 0.20533573 0.19557101 -0.03936168]]
db = [[-0.11459244]]
relu:
dA_prev = [[ 0.44090989 0. ]
[ 0.37883606 0. ]
[-0.2298228 0. ]]
dW = [[ 0.89027649 0.74742835 -0.20957978]]
db = [[-0.41675785]]
###Markdown
**Expected output with sigmoid:** dA_prev [[ 0.11017994 0.01105339] [ 0.09466817 0.00949723] [-0.05743092 -0.00576154]] dW [[ 0.20533573 0.19557101 -0.03936168]] db [[-0.11459244]] **Expected output with relu:** dA_prev [[ 0.44090989 0. ] [ 0.37883606 0. ] [-0.2298228 0. ]] dW [[ 0.89027649 0.74742835 -0.20957978]] db [[-0.41675785]] 6 - Update ParametersIn this section you will update the parameters of the model, using gradient descent: $$ W^{[1]} = W^{[1]} - \alpha \text{ } dW^{[1]} \tag{16}$$$$ b^{[1]} = b^{[1]} - \alpha \text{ } db^{[1]} \tag{17}$$$$ W^{[2]} = W^{[2]} - \alpha \text{ } dW^{[2} \tag{16}$$$$ b^{[2]} = b^{[2]} - \alpha \text{ } db^{[2]} \tag{17}$$where $\alpha$ is the learning rate. After computing the updated parameters, store them in the parameters dictionary. **Exercise**: Implement `update_parameters()` to update your parameters using gradient descent.**Instructions**:Update parameters using gradient descent.
###Code
# GRADED FUNCTION: update_parameters
def update_parameters(parameters, grads, learning_rate):
"""
Update parameters using gradient descent
Arguments:
parameters -- python dictionary containing your parameters
grads -- python dictionary containing your gradients, output of L_model_backward
Returns:
parameters -- python dictionary containing your updated parameters
parameters["W" + str(l)] = ...
parameters["b" + str(l)] = ...
"""
# Update rule for each parameter. Use a for loop.
### START CODE HERE ### (≈ 4 lines of code)
for key in parameters:
parameters[key] = parameters[key] - (learning_rate * grads["d" + str(key)])
### END CODE HERE ###
return parameters
np.random.seed(2)
W1 = np.random.randn(3,4)
b1 = np.random.randn(3,1)
W2 = np.random.randn(1,3)
b2 = np.random.randn(1,1)
parameters = {"W1": W1,
"b1": b1,
"W2": W2,
"b2": b2}
np.random.seed(3)
dW1 = np.random.randn(3,4)
db1 = np.random.randn(3,1)
dW2 = np.random.randn(1,3)
db2 = np.random.randn(1,1)
grads = {"dW1": dW1,
"db1": db1,
"dW2": dW2,
"db2": db2}
parameters = update_parameters(parameters, grads, 0.1)
print ("W1 = "+ str(parameters["W1"]))
print ("b1 = "+ str(parameters["b1"]))
print ("W2 = "+ str(parameters["W2"]))
print ("b2 = "+ str(parameters["b2"]))
###Output
W1 = [[-0.59562069 -0.09991781 -2.14584584 1.82662008]
[-1.76569676 -0.80627147 0.51115557 -1.18258802]
[-1.0535704 -0.86128581 0.68284052 2.20374577]]
b1 = [[-0.04659241]
[-1.28888275]
[ 0.53405496]]
W2 = [[-0.55569196 0.0354055 1.32964895]]
b2 = [[-0.84610769]]
###Markdown
**Expected Output**: W1 [[-0.59562069 -0.09991781 -2.14584584 1.82662008] [-1.76569676 -0.80627147 0.51115557 -1.18258802] [-1.0535704 -0.86128581 0.68284052 2.20374577]] b1 [[-0.04659241] [-1.28888275] [ 0.53405496]] W2 [[-0.55569196 0.0354055 1.32964895]] b2 [[-0.84610769]] 7 - ConclusionCongrats on implementing all the functions required for building a deep neural network! We know it was a long assignment but going forward it will only get better. The next part of the assignment is easier. Part 2:In the next part you will put all these together to build a two-layer neural networks for image classification.
###Code
%matplotlib inline
plt.rcParams['figure.figsize'] = (5.0, 4.0) # set default size of plots
plt.rcParams['image.interpolation'] = 'nearest'
plt.rcParams['image.cmap'] = 'gray'
%load_ext autoreload
%autoreload 2
np.random.seed(1)
###Output
_____no_output_____
###Markdown
Dataset **Problem Statement**: You are given a dataset ("data/train_catvnoncat.h5", "data/test_catvnoncat.h5") containing: - a training set of m_train images labelled as cat (1) or non-cat (0) - a test set of m_test images labelled as cat and non-cat - each image is of shape (num_px, num_px, 3) where 3 is for the 3 channels (RGB).Let's get more familiar with the dataset. Load the data by completing the function and run the cell below.
###Code
def load_data(train_file, test_file):
# Load the training data
train_dataset = h5py.File(train_file, 'r')
# Separate features(x) and labels(y) for training set
train_set_x_orig = np.array(train_dataset['train_set_x'])
train_set_y_orig = np.array(train_dataset['train_set_y'])
# Load the test data
test_dataset = h5py.File(test_file, 'r')
# Separate features(x) and labels(y) for training set
test_set_x_orig = np.array(test_dataset['test_set_x'])
test_set_y_orig = np.array(test_dataset['test_set_y'])
classes = np.array(test_dataset["list_classes"][:]) # the list of classes
train_set_y_orig = train_set_y_orig.reshape((1, train_set_y_orig.shape[0]))
test_set_y_orig = test_set_y_orig.reshape((1, test_set_y_orig.shape[0]))
return train_set_x_orig, train_set_y_orig, test_set_x_orig, test_set_y_orig, classes
train_file="data/train_catvnoncat.h5"
test_file="data/test_catvnoncat.h5"
train_x_orig, train_y, test_x_orig, test_y, classes = load_data(train_file, test_file)
###Output
_____no_output_____
###Markdown
The following code will show you an image in the dataset. Feel free to change the index and re-run the cell multiple times to see other images.
###Code
# Example of a picture
index = 10
plt.imshow(train_x_orig[index])
print ("y = " + str(train_y[0,index]) + ". It's a " + classes[train_y[0,index]].decode("utf-8") + " picture.")
# Explore your dataset
m_train = train_x_orig.shape[0]
num_px = train_x_orig.shape[1]
m_test = test_x_orig.shape[0]
print ("Number of training examples: " + str(m_train))
print ("Number of testing examples: " + str(m_test))
print ("Each image is of size: (" + str(num_px) + ", " + str(num_px) + ", 3)")
print ("train_x_orig shape: " + str(train_x_orig.shape))
print ("train_y shape: " + str(train_y.shape))
print ("test_x_orig shape: " + str(test_x_orig.shape))
print ("test_y shape: " + str(test_y.shape))
###Output
Number of training examples: 209
Number of testing examples: 50
Each image is of size: (64, 64, 3)
train_x_orig shape: (209, 64, 64, 3)
train_y shape: (1, 209)
test_x_orig shape: (50, 64, 64, 3)
test_y shape: (1, 50)
###Markdown
As usual, you reshape and standardize the images before feeding them to the network. Figure 1: Image to vector conversion.
###Code
# Reshape the training and test examples
train_x_flatten = train_x_orig.reshape(train_x_orig.shape[0], -1).T # The "-1" makes reshape flatten the remaining dimensions
test_x_flatten = test_x_orig.reshape(test_x_orig.shape[0], -1).T
# Standardize data to have feature values between 0 and 1.
train_x = train_x_flatten/255.
test_x = test_x_flatten/255.
print ("train_x's shape: " + str(train_x.shape))
print ("test_x's shape: " + str(test_x.shape))
###Output
train_x's shape: (12288, 209)
test_x's shape: (12288, 50)
###Markdown
3 - Architecture of your modelNow that you are familiar with the dataset, it is time to build a deep neural network to distinguish cat images from non-cat images. 2-layer neural network Figure 2: 2-layer neural network. The model can be summarized as: ***INPUT -> LINEAR -> RELU -> LINEAR -> SIGMOID -> OUTPUT***. Detailed Architecture of figure 2:- The input is a (64,64,3) image which is flattened to a vector of size $(12288,1)$. - The corresponding vector: $[x_0,x_1,...,x_{12287}]^T$ is then multiplied by the weight matrix $W^{[1]}$ of size $(n^{[1]}, 12288)$.- You then add a bias term and take its relu to get the following vector: $[a_0^{[1]}, a_1^{[1]},..., a_{n^{[1]}-1}^{[1]}]^T$.- You multiply the resulting vector by $W^{[2]}$ and add your intercept (bias). - Finally, you take the sigmoid of the result. If it is greater than 0.5, you classify it to be a cat. General methodologyAs usual you will follow the Deep Learning methodology to build the model: 1. Initialize parameters / Define hyperparameters 2. Loop for num_iterations: a. Forward propagation b. Compute loss function c. Backward propagation d. Update parameters (using parameters, and grads from backprop) 4. Use trained parameters to predict labelsLet's now implement those the model! **Question**: Use the helper functions you have implemented in the previous assignment to build a 2-layer neural network with the following structure: *LINEAR -> RELU -> LINEAR -> SIGMOID*. The functions you may need and their inputs are:```pythondef initialize_parameters(n_x, n_h, n_y): ... return parameters def linear_activation_forward(A_prev, W, b, activation): ... return A, cachedef compute_loss(AL, Y): ... return lossdef linear_activation_backward(dA, cache, activation): ... return dA_prev, dW, dbdef update_parameters(parameters, grads, learning_rate): ... return parameters```
###Code
### CONSTANTS DEFINING THE MODEL ####
n_x = 12288 # num_px * num_px * 3
n_h = 7
n_y = 1
layers_dims = (n_x, n_h, n_y)
def two_layer_model(X, Y, layers_dims, learning_rate = 0.0075, num_iterations = 3000, print_loss=False):
"""
Implements a two-layer neural network: LINEAR->RELU->LINEAR->SIGMOID.
Arguments:
X -- input data, of shape (n_x, number of examples)
Y -- true "label" vector (containing 0 if cat, 1 if non-cat), of shape (1, number of examples)
layers_dims -- dimensions of the layers (n_x, n_h, n_y)
num_iterations -- number of iterations of the optimization loop
learning_rate -- learning rate of the gradient descent update rule
print_loss -- If set to True, this will print the loss every 100 iterations
Returns:
parameters -- a dictionary containing W1, W2, b1, and b2
"""
np.random.seed(1)
grads = {}
losses = [] # to keep track of the loss
m = X.shape[1] # number of examples
(n_x, n_h, n_y) = layers_dims
# Initialize parameters dictionary, by calling one of the functions you'd previously implemented
### START CODE HERE ### (≈ 1 line of code)
parameters = initialize_parameters(n_x, n_h, n_y)
### END CODE HERE ###
# Get W1, b1, W2 and b2 from the dictionary parameters.
W1 = parameters["W1"]
b1 = parameters["b1"]
W2 = parameters["W2"]
b2 = parameters["b2"]
# Loop (gradient descent)
for i in range(0, num_iterations):
# Forward propagation: LINEAR -> RELU -> LINEAR -> SIGMOID. Inputs: "X, W1, b1, W2, b2". Output: "A1, cache1, A2, cache2".
### START CODE HERE ### (≈ 2 lines of code)
A1, cache1 = linear_activation_forward(X, W1, b1, "relu")
A2, cache2 = linear_activation_forward(A1, W2, b2, "sigmoid")
### END CODE HERE ###
# Compute loss
### START CODE HERE ### (≈ 1 line of code)
loss = compute_loss(A2, Y)
### END CODE HERE ###
# Initializing backward propagation
dA2 = - (np.divide(Y, A2) - np.divide(1 - Y, 1 - A2))/m
# Backward propagation. Inputs: "dA2, cache2, cache1". Outputs: "dA1, dW2, db2; also dA0 (not used), dW1, db1".
### START CODE HERE ### (≈ 2 lines of code)
dA1, dW2, db2 = linear_activation_backward(dA2, cache2, "sigmoid")
dA0, dW1, db1 = linear_activation_backward(dA1, cache1, "relu")
### END CODE HERE ###
# Set grads['dWl'] to dW1, grads['db1'] to db1, grads['dW2'] to dW2, grads['db2'] to db2
### START CODE HERE ### (≈ 4 lines of code)
grads['dW1'] = dW1
grads['db1'] = db1
grads['dW2'] = dW2
grads['db2'] = db2
### END CODE HERE ###
# Update parameters.
### START CODE HERE ### (approx. 1 line of code)
parameters = update_parameters(parameters, grads, learning_rate)
### END CODE HERE ###
# Retrieve W1, b1, W2, b2 from parameters
W1 = parameters["W1"]
b1 = parameters["b1"]
W2 = parameters["W2"]
b2 = parameters["b2"]
# Print the loss every 100 training example
if print_loss and i % 100 == 0:
print("Loss after iteration {}: {}".format(i, np.squeeze(loss)))
if print_loss and i % 100 == 0:
losses.append(loss)
# plot the loss
plt.plot(np.squeeze(losses))
plt.ylabel('loss')
plt.xlabel('iterations (per tens)')
plt.title("Learning rate =" + str(learning_rate))
plt.show()
return parameters
parameters = two_layer_model(train_x, train_y, layers_dims = (n_x, n_h, n_y), learning_rate=0.003, num_iterations = 7000, print_loss=True)
###Output
Loss after iteration 0: 0.69304973566
Loss after iteration 100: 0.66358362054
Loss after iteration 200: 0.648272816823
Loss after iteration 300: 0.644370511753
Loss after iteration 400: 0.639397427608
Loss after iteration 500: 0.632374537296
Loss after iteration 600: 0.6222656789
Loss after iteration 700: 0.608749048358
Loss after iteration 800: 0.593317668528
Loss after iteration 900: 0.577051173035
Loss after iteration 1000: 0.559815202387
Loss after iteration 1100: 0.54260614129
Loss after iteration 1200: 0.524435258075
Loss after iteration 1300: 0.50629684946
Loss after iteration 1400: 0.488534059566
Loss after iteration 1500: 0.470691257416
Loss after iteration 1600: 0.452698521786
Loss after iteration 1700: 0.430112987407
Loss after iteration 1800: 0.40371365539
Loss after iteration 1900: 0.379037148229
Loss after iteration 2000: 0.355524892053
Loss after iteration 2100: 0.334077833925
Loss after iteration 2200: 0.31377615142
Loss after iteration 2300: 0.295163988192
Loss after iteration 2400: 0.277980886773
Loss after iteration 2500: 0.26172341494
Loss after iteration 2600: 0.24673638906
Loss after iteration 2700: 0.232723655169
Loss after iteration 2800: 0.219602225882
Loss after iteration 2900: 0.207130090975
Loss after iteration 3000: 0.1956742445
Loss after iteration 3100: 0.184661640146
Loss after iteration 3200: 0.173723977661
Loss after iteration 3300: 0.163918461958
Loss after iteration 3400: 0.154874149793
Loss after iteration 3500: 0.145669560808
Loss after iteration 3600: 0.137200381502
Loss after iteration 3700: 0.129435190953
Loss after iteration 3800: 0.122515828644
Loss after iteration 3900: 0.11610318636
Loss after iteration 4000: 0.110072191657
Loss after iteration 4100: 0.10406774094
Loss after iteration 4200: 0.0984590798725
Loss after iteration 4300: 0.0935180956028
Loss after iteration 4400: 0.0890586140023
Loss after iteration 4500: 0.0847087401318
Loss after iteration 4600: 0.0808851905079
Loss after iteration 4700: 0.0774511672646
Loss after iteration 4800: 0.0739222258769
Loss after iteration 4900: 0.070827789301
Loss after iteration 5000: 0.0679165909286
Loss after iteration 5100: 0.0651693765764
Loss after iteration 5200: 0.062481603939
Loss after iteration 5300: 0.0599379765221
Loss after iteration 5400: 0.0575760866184
Loss after iteration 5500: 0.0553193553812
Loss after iteration 5600: 0.0531034948389
Loss after iteration 5700: 0.0510428466166
Loss after iteration 5800: 0.0490311649651
Loss after iteration 5900: 0.0470733380646
Loss after iteration 6000: 0.045215782521
Loss after iteration 6100: 0.0434024427312
Loss after iteration 6200: 0.0416500470376
Loss after iteration 6300: 0.0399726228339
Loss after iteration 6400: 0.0383499390173
Loss after iteration 6500: 0.0368139336141
Loss after iteration 6600: 0.0353062856631
Loss after iteration 6700: 0.0338848414523
Loss after iteration 6800: 0.032537506473
Loss after iteration 6900: 0.0312628723581
###Markdown
**Expected Output**: **Loss after iteration 0** 0.6930497356599888 **Loss after iteration 100** 0.6464320953428849 **...** ... **Loss after iteration 2400** 0.048554785628770206 Good thing you built a vectorized implementation! Otherwise it might have taken 10 times longer to train this.Now, you can use the trained parameters to classify images from the dataset. ***Exercise:*** - Implement the forward function- Implement the predict function below to make prediction on test_images
###Code
def two_layer_forward(X, parameters):
"""
Implement forward propagation for the [LINEAR->RELU]*(L-1)->LINEAR->SIGMOID computation
Arguments:
X -- data, numpy array of shape (input size, number of examples)
parameters -- output of initialize_parameters_deep()
Returns:
AL -- last post-activation value
caches -- list of caches containing:
every cache of linear_relu_forward() (there are L-1 of them, indexed from 0 to L-2)
the cache of linear_sigmoid_forward() (there is one, indexed L-1)
"""
caches = []
A = X
# Implement LINEAR -> RELU. Add "cache" to the "caches" list.
### START CODE HERE ### (approx. 3 line of code)
W1, b1 = parameters["W1"], parameters["b1"]
A1, cache1 = linear_activation_forward(A, W1, b1, "relu")
caches.append(cache1)
### END CODE HERE ###
# Implement LINEAR -> SIGMOID. Add "cache" to the "caches" list.
### START CODE HERE ### (approx. 3 line of code)
W2, b2 = parameters["W2"], parameters["b2"]
A2, cache2 = linear_activation_forward(A1, W2, b2, "sigmoid")
caches.append(cache2)
### END CODE HERE ###
assert(A2.shape == (1,X.shape[1]))
return A2, caches
def predict(X, y, parameters):
"""
This function is used to predict the results of a L-layer neural network.
Arguments:
X -- data set of examples you would like to label
parameters -- parameters of the trained model
Returns:
p -- predictions for the given dataset X
"""
m = X.shape[1]
n = len(parameters) // 2 # number of layers in the neural network
p = np.zeros((1,m))
# Forward propagation
### START CODE HERE ### (≈ 1 lines of code)
probas, caches = two_layer_forward(X, parameters)
### END CODE HERE ###
# convert probas to 0/1 predictions
for i in range(0, probas.shape[1]):
### START CODE HERE ### (≈ 4 lines of code)
if(probas[0][i] > 0.5):
p[0][i] = 1
else:
p[0][i] = 0
### END CODE HERE ###
print("Accuracy: " + str(float(np.sum((p == y)/float(m)))))
return p
predictions_train = predict(train_x, train_y, parameters)
predictions_test = predict(test_x, test_y, parameters)
###Output
Accuracy: 0.76
###Markdown
***Exercise:***Identify the hyperparameters in the model and For each hyperparameter- Briefly explain its role- Explore a range of values and describe their impact on (a) training loss and (b) test accuracy- Report the best hyperparameter value found.Note: Provide your results and explanations in the report for this question. **Hyperparameters**The hyperparameters are:1. Learning rate - It is used for updating the parameters of the neural network that is the weights and biases of the neural network. It controls the amount of update that needs to take place so that we are able to reach the minima of the loss function.2. Epochs - It represents the number of times the network sees the data and adjusts its parameters for optimal learning.**Values of Hyperparameters tried:**1. Learning rate = 0.005, Epochs = 2000, Training loss = 0.187, Testing accuracy: 70%2. Learning rate = 0.005, Epochs = 3000, Training loss = 0.073, Testing accuracy: 72%3. Learning rate = 0.005, Epochs = 4000, Training loss = 0.036, Testing accuracy: 72%4. Learning rate = 0.001, Epochs = 2000, Training loss = 0.617, Testing accuracy: 34%5. Learning rate = 0.001, Epochs = 3000, Training loss = 0.56, Testing accuracy: 34%6. Learning rate = 0.001, Epochs = 4000, Training loss = 0.5, Testing accuracy: 34%7. Learning rate = 0.01, Epochs = 2000, Training loss = 0.05, Testing accuracy: 72%8. Learning rate = 0.01, Epochs = 3000, Training loss = 0.01, Testing accuracy: 72%9. Learning rate = 0.01, Epochs = 4000, Training loss = 0.008, Testing accuracy: 72%10. Learning rate = 0.05, Epochs = 2000, Training loss = 0.58, Testing accuracy: 46%11. Learning rate = 0.01, Epochs = 3000, Training loss = 0.36, Testing accuracy: 56%12. Learning rate = 0.03, Epochs = 2000, Training loss = 0.0058, Testing accuracy: 72%13. Learning rate = 0.03, Epochs = 3000, Training loss = 0.0024, Testing accuracy: 72%14. Learning rate = 0.003, Epochs = 4000, Training loss = 0.11, Testing accuracy: 76%15. Learning rate = 0.003, Epochs = 7000, Training loss = 0.047, Testing accuracy: 76%16. Learning rate = 0.003, Epochs = 8000, Training loss = 0.021, Testing accuracy: 74%17. Learning rate = 0.002, Epochs = 4000, Training loss = 0.24, Testing accuracy: 76%18. Learning rate = 0.002, Epochs = 6000, Training loss = 0.113, Testing accuracy: 74%19. Learning rate = 0.002, Epochs = 8000, Training loss = 0.06, Testing accuracy: 74%18. Learning rate = 0.0025, Epochs = 6000, Training loss = 0.07, Testing accuracy: 74%19. Learning rate = 0.0025, Epochs = 8000, Training loss = 0.03, Testing accuracy: 72%**Optimal hyperparameters found**1. Learning rate = 0.0032. Epochs = 7000 Results AnalysisFirst, let's take a look at some images the 2-layer model labeled incorrectly. This will show a few mislabeled images.
###Code
def print_mislabeled_images(classes, X, y, p):
"""
Plots images where predictions and truth were different.
X -- dataset
y -- true labels
p -- predictions
"""
a = p + y
mislabeled_indices = np.asarray(np.where(a == 1))
plt.rcParams['figure.figsize'] = (40.0, 40.0) # set default size of plots
num_images = len(mislabeled_indices[0])
for i in range(num_images):
index = mislabeled_indices[1][i]
plt.subplot(2, num_images, i + 1)
plt.imshow(X[:,index].reshape(64,64,3), interpolation='nearest')
plt.axis('off')
plt.title("Prediction: " + classes[int(p[0,index])].decode("utf-8") + " \n Class: " + classes[y[0,index]].decode("utf-8"))
print_mislabeled_images(classes, test_x, test_y, predictions_test)
###Output
_____no_output_____
###Markdown
***Exercise:*** Identify a few types of images that tends to perform poorly on the model **Answer**The model performs poorly when the cat is at certain angle or rotated at some angle, which makes it classify it as a non-cat class. Now, lets use the same architecture to predict sentiment of movie reviews. In this section, most of the implementation is already provided. The exercises are mainly to understand what the workflow is when handling the text data.
###Code
import re
###Output
_____no_output_____
###Markdown
Dataset **Problem Statement**: You are given a dataset ("train_imdb.txt", "test_imdb.txt") containing: - a training set of m_train reviews - a test set of m_test reviews - the labels for the training examples are such that the first 50% belong to class 1 (positive) and the rest 50% of the data belong to class 0(negative) Let's get more familiar with the dataset. Load the data by completing the function and run the cell below.
###Code
def load_data(train_file, test_file):
train_dataset = []
test_dataset = []
# Read the training dataset file line by line
for line in open(train_file, 'r'):
train_dataset.append(line.strip())
for line in open(test_file, 'r'):
test_dataset.append(line.strip())
return train_dataset, test_dataset
train_file = "data/train_imdb.txt"
test_file = "data/test_imdb.txt"
train_dataset, test_dataset = load_data(train_file, test_file)
# This is just how the data is organized. The first 50% data is positive and the rest 50% is negative for both train and test splits.
y = [1 if i < len(train_dataset)*0.5 else 0 for i in range(len(train_dataset))]
###Output
_____no_output_____
###Markdown
As usual, lets check our dataset
###Code
# Example of a review
index = 10
print(train_dataset[index])
print ("y = " + str(y[index]))
# Explore your dataset
m_train = len(train_dataset)
m_test = len(test_dataset)
print ("Number of training examples: " + str(m_train))
print ("Number of testing examples: " + str(m_test))
###Output
Number of training examples: 1001
Number of testing examples: 201
###Markdown
Pre-Processing From the example review, you can see that the raw data is really noisy! This is generally the case with the text data. Hence, Preprocessing the raw input and cleaning the text is essential. Please run the code snippet provided below.**Exercise**: Explain what pattern the model is trying to capture using re.compile in your report. **Answer**1. re.compile() removes special characters like ', . " etc and makes all characters in lowercase. It is learning properties from words.
###Code
REPLACE_NO_SPACE = re.compile("(\.)|(\;)|(\:)|(\!)|(\')|(\?)|(\,)|(\")|(\()|(\))|(\[)|(\])|(\d+)")
REPLACE_WITH_SPACE = re.compile("(<br\s*/><br\s*/>)|(\-)|(\/)")
NO_SPACE = ""
SPACE = " "
def preprocess_reviews(reviews):
reviews = [REPLACE_NO_SPACE.sub(NO_SPACE, line.lower()) for line in reviews]
reviews = [REPLACE_WITH_SPACE.sub(SPACE, line) for line in reviews]
return reviews
train_dataset_clean = preprocess_reviews(train_dataset)
test_dataset_clean = preprocess_reviews(test_dataset)
# Example of a clean review
index = 10
print(train_dataset_clean[index])
print ("y = " + str(y[index]))
###Output
i liked the film some of the action scenes were very interesting tense and well done i especially liked the opening scene which had a semi truck in it a very tense action scene that seemed well done some of the transitional scenes were filmed in interesting ways such as time lapse photography unusual colors or interesting angles also the film is funny is several parts i also liked how the evil guy was portrayed too id give the film an out of
y = 1
###Markdown
Vectorization Now lets create a feature vector for our reviews based on a simple bag of words model. So, given an input text, we need to create a numerical vector which is simply the vector of word counts for each word of the vocabulary. Run the code below to get the feature representation.
###Code
from sklearn.feature_extraction.text import CountVectorizer
cv = CountVectorizer(binary=True, stop_words="english", max_features=2000)
cv.fit(train_dataset_clean)
X = cv.transform(train_dataset_clean)
X_test = cv.transform(test_dataset_clean)
###Output
_____no_output_____
###Markdown
CountVectorizer provides a sparse feature representation by default which is reasonable because only some words occur in individual example. However, for training neural network models, we generally use a dense representation vector.
###Code
X = np.array(X.todense()).astype(float)
X_test = np.array(X_test.todense()).astype(float)
y = np.array(y)
###Output
_____no_output_____
###Markdown
Model
###Code
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
X_train, X_val, y_train, y_val = train_test_split(
X, y, train_size = 0.80
)
# This is just to correct the shape of the arrays as required by the two_layer_model
X_train = X_train.T
X_val = X_val.T
y_train = y_train.reshape(1,-1)
y_val = y_val.reshape(1,-1)
### CONSTANTS DEFINING THE MODEL ####
n_x = X_train.shape[0]
n_h = 200
n_y = 1
layers_dims = (n_x, n_h, n_y)
###Output
_____no_output_____
###Markdown
We will use the same two layer model that you completed in the previous section for training.
###Code
parameters = two_layer_model(X_train, y_train, layers_dims = (n_x, n_h, n_y), learning_rate=0.05, num_iterations = 3000, print_loss=True)
###Output
Loss after iteration 0: 0.693079416169
Loss after iteration 100: 0.686569463673
Loss after iteration 200: 0.653746782902
Loss after iteration 300: 0.513637888328
Loss after iteration 400: 0.320195643458
Loss after iteration 500: 0.201724848093
Loss after iteration 600: 0.134819976219
Loss after iteration 700: 0.0948401309064
Loss after iteration 800: 0.0695863625855
Loss after iteration 900: 0.0528916264559
Loss after iteration 1000: 0.041486399968
Loss after iteration 1100: 0.0334640145333
Loss after iteration 1200: 0.027650556749
Loss after iteration 1300: 0.0233134006421
Loss after iteration 1400: 0.0199911510628
Loss after iteration 1500: 0.0173876238095
Loss after iteration 1600: 0.0153066762094
Loss after iteration 1700: 0.0136146620512
Loss after iteration 1800: 0.0122180889721
Loss after iteration 1900: 0.0110502249368
Loss after iteration 2000: 0.010062214655
Loss after iteration 2100: 0.00921766338755
Loss after iteration 2200: 0.00848910281476
Loss after iteration 2300: 0.00785537317867
Loss after iteration 2400: 0.00730000461864
Loss after iteration 2500: 0.00681003094208
Loss after iteration 2600: 0.00637506645976
Loss after iteration 2700: 0.00598679650114
Loss after iteration 2800: 0.00563846787071
Loss after iteration 2900: 0.0053244913391
###Markdown
Predict the review for our movies!
###Code
predictions_train = predict(X_train, y_train, parameters)
predictions_val = predict(X_val, y_val, parameters)
###Output
Accuracy: 0.850746268657
###Markdown
Results AnalysisLet's take a look at some examples the 2-layer model labeled incorrectly
###Code
def print_mislabeled_reviews(X, y, p):
"""
Plots images where predictions and truth were different.
X -- dataset
y -- true labels
p -- predictions
"""
a = p + y
mislabeled_indices = np.asarray(np.where(a == 1))
plt.rcParams['figure.figsize'] = (40.0, 40.0) # set default size of plots
num_reviews = len(mislabeled_indices[0])
for i in range(num_reviews):
index = mislabeled_indices[1][i]
print((" ").join(cv.inverse_transform(X[index])[0]))
print("Prediction: " + str(int(p[0,index])) + " \n Class: " + str(y[0,index]))
print_mislabeled_reviews(X_val.T, y_val, predictions_val)
###Output
actors attempt beauty believable big bit charismatic claims definitely delivery did didnt disappointing disaster entertained fact film fine group job line looked lost miss offensive performance playing plays plot project recommend rent scenes screen seen strong talent wish writing
Prediction: 0
Class: 1
acting add annoying bad change character dicaprio did director does eyes film filmmakers films glad going good great half hand hardly impressive just kate learned lesson love mean million movie opinion oscar performance possible really romance romantic second ship shouldnt single sit stories story sure talented think thinking time times titanic try watching win wonderful wont worst
Prediction: 0
Class: 1
anna appearance away bad better bible big black blue book boys build capture cat catch chaos charles child city comic connected cops cult deal didnt doesnt earth edge exactly extreme far favorite fictional fine finished followed fond form fun gang gas genius giant god going good got government green guy guys half happening happens hard havent having heroes hey hospital include japanese join just kind know known like liked line looks lot make match monster movie mysterious naked named names new nice oh order paid past people place places police power problem project puts quickly red remains right seconds seeing series set sexy soon sorts special starts story taking thanks thats theres theyre thing tom unfortunately use using villains violence want whos woman world yeah year youd
Prediction: 0
Class: 1
able action add admit adult ahead anna bit black boy characters cinematography come deserves disappointed disappointment does dont doubt downright elements end entirely expect expecting fear film forget genre girls guys hopes imagination instead intense knew leave lesbian level like little looking lot love managed memorable mid movie ok performances play pleasure prepared read real realized really received romance school secret shot shown soon stars story straight sudden teenagers theyre think time times trying unexpected watching way white women wont wrong years youll
Prediction: 0
Class: 1
actor actually adults bringing calls cast child children doing era eyes famous focus fun given guess guy history host interesting john julia kenneth kind king like martin movie natural news park police provided question really say seeing short shouldnt simply smith sort story thought true version voice voices woman work worth wouldnt young
Prediction: 0
Class: 1
based got involved movie moving mystery oscar review script slow star
Prediction: 1
Class: 0
actually ahead better box budget burning character charlie come damn development did didnt dont enjoy eye fake far film flick forced genre good guinea guts hand hear heard horror hours interested just know like listen looks lot low making men minutes movie naturally offer painful pretty really recommend say scene scenes second seen set sharp short simply snuff story think thought throwing told torture trying ultimately unless various watching ways went woman worst
Prediction: 0
Class: 1
cause early effort government heavy past people problems production propaganda short spending sudden time truly using war window
Prediction: 1
Class: 0
ability acting actor ages anti better cheap cinema complete confused day decent direction disappointed disappointment does dont dull dvd explanation film finally finding flat gate good got great guess intriguing kind lead like liked love main meet mouth movie performance plot poor premise remind required result rip saw say store story tell thought took tv wanted way week written
Prediction: 1
Class: 0
action adventure adventures bad camp character characters check crew decided doc elements familiar fan fans feel feeling film good hero heroes im james jones just know long lot major minutes movie movies music number ones promise provided really resulting savage say seeing somewhat spirit star thats theres throw time trying unfortunate way
Prediction: 0
Class: 1
better body cast central cinema come coming comments company computer decides die entertaining exist failed fall fan film films genius gets getting going great hard hey highly hollywood house idea interested judging latest lesson lessons like make making man member money movie movies near potential premise puts review reviews role scenes seen set shock soon stupid taken takes type unless unrelated wish wow writer writing yes zero
Prediction: 1
Class: 0
admit almighty attempt big bruce carrey cast cheesy comedy dont end enjoyable fan feel funny gets gone havent help hilarious ill im jim just know let light like movie movies music note poor positive really rest reviews saying seen shows somewhat start steve thinking want writers youre
Prediction: 0
Class: 1
action age body brain building certainly computer crazy damme daughter dead entertaining especially fan fi fights folks genius goes going goldberg good government guess hes humor just keeps king lame later latest like manages mean named new original particularly perfect power pretty pro reason run sci sequel shoot site snake soldiers sort step super takes thriller train usual van war white working wrong year years youre
Prediction: 1
Class: 0
acting animals best better die dont entire episode episodes funny good horrible ice just killing know life like movie obviously plot pro problem really remember right scene scenes season second series shocking suspense think torture turns victims watch women wonderful worst
Prediction: 0
Class: 1
bunch doesnt feel got laugh laughed left like loud make masterpiece movie ok purpose smile times viewer worth
Prediction: 0
Class: 1
acting away beautifully biggest burt came character drinking fact failure fast fell general help hoping job movie movies night notice played promising real right screen single state thats walk way
Prediction: 1
Class: 0
charlie dont eye fake film final harder hot im know like look looks real said say scene scenes sure tell thing truth
Prediction: 0
Class: 1
actually ago american begin begins big bring buy century circumstances couple does doesnt effects emotional flicks follows happen highly home house husband impact john life like man masterpiece mysterious old outside plot recommended simple special story strange supposedly things turn unknown went woman world
Prediction: 0
Class: 1
absolutely add bad best better boat book brought cases classic clear cliché close course critics deserves didnt disappointed exactly excitement family felt field film finally giving grew hard hear hero heroes home ill im instead know latest like line mind missing musical names nature non offensive old particularly past poor professional race real reality reviewer ridiculous right rock sadly said scene scenes score sense shot shots shows smile sound spot starting supposed taking talking theres theyve thrill time took town versions water wonderful years yes
Prediction: 1
Class: 0
ability able accident action actresses actually aspect away bad believe better bit blood brothers cause cgi charlie crap crime cut days deal death disturbing does doesnt dont effects especially eyes fact fake favorite film films footage forget funny happens hope horror im instead leaving like look lot make makers making marry money movie movies overall people plot point porn probably pull rape rating real saying says scene scenes seen series shocking snuff sound stand stars sucked sucks super supposed sure talent talking thing thinking time tried visual visuals want wanted wasnt watch wouldnt
Prediction: 0
Class: 1
achieve acting approach art artistic background box brief cast cheap cinema close come cons considered consists contemporary country dealing deals deserves director fact fan fit good half hard history hope hot huge job just knows like make manage masterpiece meant media members money movie naked near office ones opinion perfect perfectly provide purpose real roles short single small success talent theatrical thing time touching tried usual waiting women word work
Prediction: 1
Class: 0
acting actors admit annoying arent art bad ball beginning best better big billy bits book calling camera case character characters cinematic come coming cons crouse cusack david definitely dialogue did didnt direct directed does doesnt dont early end ending entertaining expecting extremely far feel film filmed films flat forth free fun game games gets getting girl going good guy half help heres hes hour house ill im inner involved isnt james john just keeps lesson let level like lindsay line lines little look looked lose mamet mantegna mark maybe mean men middle mind minutes moves movie narration nature new ones opening pick play precious pretty problem quality questions read reading real realize really result ring roll room scene second shes sort sound sounds speaking standard start stick story strange stuff supposed theatre theyre things true want wants watch way weird whats words work wouldnt write
Prediction: 0
Class: 1
accept ago army away bad begins body bucks budget chase comes couple dolph door energy especially exist explained feel feeling fight fighting films flash flick follow forward goes good happens hell human idea ideas involved isnt just key lacks like long looks low lukas make man master member merely middle movie movies needless new order place plays potential previous satan say scene scenes secret sense sort stars story study sucks supposed sure takes theres thrown time underground wish wont years york youll
Prediction: 1
Class: 0
absolutely acted art audience bad bar beginning came chinese come coming comments course deep didnt director doing drawn end ending entertaining essentially experience faces fact fantastic far feel festival film final following forget fresh fun gonna government half happy hard hidden hollywood hour hours im immediately incredibly intelligent intriguing judging just land late life likable long looked lot loved make making match meaning mention natural new number pain painful point post probably problem promising reading really reason reviews right russian said saw say sense sharp simply society sounds spent started state talking thank theatre thought time took try utter utterly view want wanted warned way week whats whatsoever words working years yes
Prediction: 1
Class: 0
actually ago bad better book church course does enjoyable familiar film forgotten forward good hadnt heard hour instantly job know laid let long minute minutes missed mr nearly overall quick read really school second sense short simply story tales thats thing time trilogy watched worked write years
Prediction: 0
Class: 1
actors ask blood cares character conclusion content crap crew damn day disturbing effort ends episode exception family fan fate gets going gore great gross hopes horror hour imagine lot mindless new performances pointless producers production reason season sense series shock stories story tend thinking utter values violence work worse
Prediction: 1
Class: 0
absolutely acting bits casting cheap close come comments completely couldve did direction edge end film gone humor intense literally little loved mediocre movie number perfect phone points read rest ring scary script second spot story thrill time years
Prediction: 0
Class: 1
ann cause characters comedy compared computer connection considered day days did dont end entertainment ex fight film films flight future george given going got hand having help hes home human including instead isnt issue just kids kill killed know lesson life like live losing lost love make making man matter maybe meets money necessary people pictures plan plays plot prior project rich school sets shows stars step street streets stupid technology tender theyre things think thrown treasure used using vote wall wants war wasnt woman work world written years young
Prediction: 0
Class: 1
actors alive based childhood documentary got kill know man mission monster movie people personality played rate real scenes set seven turn used women work
Prediction: 0
Class: 1
film forgotten late little long makers money movie night present subtle time todays tv
Prediction: 0
Class: 1
|
notebooks/dataprep/01a-UniProtProtein.ipynb | ###Markdown
UniProt Viral and Host Protein Data**[Work in progress]**This notebook downloads and standardizes viral and host protein data from UniProt for ingestion into the Knowledge Graph.Data source: [UniProt](https://www.uniprot.org/)Authors: Peter Rose ([email protected])
###Code
import os
import re
import hashlib
import urllib
import pandas as pd
import numpy as np
from pathlib import Path
pd.options.display.max_rows = None # display all rows
pd.options.display.max_columns = None # display all columns
NEO4J_IMPORT = Path(os.getenv('NEO4J_IMPORT'))
print(NEO4J_IMPORT)
###Output
/Users/peter/Library/Application Support/com.Neo4j.Relate/data/dbmss/dbms-8bf637fc-0d20-4d9f-9c6f-f7e72e92a4da/import
###Markdown
Get list of organisms in the Knowledge Graph
###Code
organisms = pd.read_csv("../../reference_data/Organism.csv", dtype=str)
# exclude organisms without an NCBI taxonomy id
organisms = organisms[organisms['id'].str.startswith('taxonomy')]
# remove CURIE
organisms['taxonomy'] = organisms['id'].apply(lambda x: x.split(':')[1])
taxonomy_ids = organisms['taxonomy'].unique()
###Output
_____no_output_____
###Markdown
Download data from UniProt
###Code
columns = 'id,entry%20name,p,sequence,length,protein%20names,reviewed,organism-id,feature(CHAIN),feature(PEPTIDE),go(biological%20process)'
dfs = list()
for taxon in taxonomy_ids:
url = f'https://www.uniprot.org/uniprot/?query=organism:{taxon}&columns={columns}&format=tab'
try:
df = pd.read_csv(url, sep='\t', dtype='str')
if df.shape[0] > 0:
print(f'Downloaded {df.shape[0]} proteins for taxonomy id {taxon}')
dfs.append(df)
else:
print(f'Downloaded 0 proteins for taxonomy id {taxon}')
except:
print(f'Downloaded 0 proteins for taxonomy id {taxon}')
unp = pd.concat(dfs)
unp.reset_index(drop=True,inplace=True)
unp.fillna('', inplace=True)
print(unp.shape)
unp.tail()
unp['reviewed'] = unp['Status'].apply(lambda s: 'True' if s == 'reviewed' else 'False')
###Output
_____no_output_____
###Markdown
Format synonymes
###Code
unp.query("Entry == 'P0DTC2'")['Protein names'].values
###Output
_____no_output_____
###Markdown
Remove terms in brackets, e.g., [Cleaved into: ...]["Spike glycoprotein (S glycoprotein) (E2) (Peplomer protein) [Cleaved into: Spike protein S1; Spike protein S2; Spike protein S2']
###Code
unp['synonymes'] = unp['Protein names'].str.replace("\\[.+\\]", "")
unp.query("Entry == 'P0DTC2'")['synonymes'].values
###Output
_____no_output_____
###Markdown
Convert synonymes to a semicolon separated list to represent these one to many relationships in a CSV file.
###Code
unp['synonymes'] = unp['synonymes'].str.replace('(', ';')
unp['synonymes'] = unp['synonymes'].str.replace(' ;', ';')
unp['synonymes'] = unp['synonymes'].str.replace(')', '')
unp['synonymes'] = unp['synonymes'].str.strip()
unp.query("Entry == 'P0DTC2'")['synonymes'].values
unp.head()
unp.query("Entry == 'P01042'")['Chain'].values
unp.query("Entry == 'P01042'")['Peptide'].values
def parse_feature_record(record, feature_type):
items = record.split(';')
feature = np.empty(5, dtype=object)
feature[0] = feature_type
for item in items:
item = item.strip()
if '..' in item:
start_end = item.split('..')
# in a few cases a '?' is used to represent an unknown start or end, check if it's a digit
if start_end[0].isdigit():
feature[1] = start_end[0]
else:
feature[1] = ''
if start_end[1].isdigit():
feature[2] = start_end[1]
else:
feature[2] = ''
elif item.startswith("/note="):
name = item[6:].replace('\"', '')
feature[3] = name
elif item.startswith("/id="):
pro_id = item[4:].replace('\"', '')
feature[4] = 'uniprot.chain:' + pro_id
return feature
def parse_features(row):
chain_features = []
if 'CHAIN' in row['Chain']:
chains = row['Chain'].split('CHAIN')
if chains[0] == '':
chains = chains[1:]
chain_features = [parse_feature_record(chain, 'CHAIN') for chain in chains]
protein_features = []
# Full-length (coding sequence) proteins are inconsistenly handled in UniProt.
# For some entries, the full-length protein is included
# in the chain features (e.g. P0DTD1), for others it's not (e.g., P01042)
# Check if full-length protein is included in chain list
full_length = False
for f in chain_features:
if f[1] == '1' and f[2] == row['Length']:
full_length = True
break
# Add entry if full-length protein is not in chain list
if not full_length:
protein_name = row['Protein names'].split('(')[0]
protein_features = [np.array(['PROTEIN','1', row['Length'], protein_name,''], dtype=object)]
peptide_features = []
if 'PEPTIDE' in row['Peptide']:
peptides = row['Peptide'].split('PEPTIDE')
if peptides[0] == '':
peptides = peptides[1:]
peptide_features = [parse_feature_record(peptide, 'PEPTIDE') for peptide in peptides]
return protein_features + chain_features + peptide_features
unp['Features'] = unp.apply(parse_features, axis=1)
unp = unp.explode('Features')
unp[['type', 'start', 'end', 'name', 'proId']] = unp.apply(lambda row: row['Features'], axis=1, result_type="expand")
###Output
_____no_output_____
###Markdown
Handle missing values
###Code
unp.fillna('', inplace=True)
unp.head(50)
###Output
_____no_output_____
###Markdown
Cleave sequences into peptides
###Code
def get_subsequence(row):
if row['start'].isdigit() and row['end'].isdigit():
start = int(row['start'])
end = int(row['end'])
sequence = row['Sequence']
return sequence[start-1: end]
else:
return ''
unp['sequence'] = unp.apply(lambda row: get_subsequence(row), axis=1)
###Output
_____no_output_____
###Markdown
Set flag if protein chain is full length
###Code
unp['fullLength'] = (unp['start'] == '1') & (unp['end'] == unp['Length'])
unp['name'] = unp['name'].str.strip()
def set_synonymes(row):
if row['fullLength']:
return row['synonymes']
else:
return row['name']
###Output
_____no_output_____
###Markdown
Cleaved protein, should not inherit synonymes from the full length protein
###Code
unp['synonymes'] = unp.apply(set_synonymes, axis=1)
unp.head(50)
unp.rename(columns={'Organism ID': 'taxonomyId','Entry': 'accession', 'Entry name': 'entryName'}, inplace=True)
###Output
_____no_output_____
###Markdown
Assign unique identifiersmd5 hashcodes for the protein sequence and CURIEs for accession and taxonomyId
###Code
unp['id'] = unp['sequence'].apply(lambda seq: 'md5:' + hashlib.md5(seq.encode()).hexdigest())
# disambiguate id by taxonomyId (same sequence for different organisms)
unp['id'] = unp['id'] + '-' + unp['taxonomyId']
unp['accession'] = 'uniprot:' + unp['accession']
unp['taxonomyId'] = 'taxonomy:' + unp['taxonomyId']
unp.query("accession == 'uniprot:P01042'")
###Output
_____no_output_____
###Markdown
Save proteins
###Code
columns = ['id', 'name', 'synonymes', 'accession', 'entryName', 'proId', 'taxonomyId','sequence',
'start', 'end', 'fullLength', 'reviewed']
unp.to_csv(NEO4J_IMPORT / '01a-UniProtProtein.csv', columns=columns, index = False)
unp.head()
###Output
_____no_output_____ |
examples/water_boiler/base_case_no_control_rods/.ipynb_checkpoints/water_boiler-checkpoint.ipynb | ###Markdown
Water BoilerThis is a description of the Water Boiler reactor - a homogeneous solution of Uranyl(14.7) Sulfate Solution in a hollow stainless steel sphere with a Beryllium Oxide reflector. The reactor was used in an experiment at LANL during the Manhattan Project. The goal of the project was to determine the critical mass of U-235 in a homogeneous solution with various reflectors used. A goal of the notebook will be to make extensive use of the Nuclear Data interface of OpenMC for all of the required data.
###Code
import openmc
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
###Output
_____no_output_____
###Markdown
Material Parameters and Calculations
###Code
# the mass density of UO2-SO4 solution in water is needed
import scipy.interpolate
def get_UranylSulfate_solution_density(pct_us=0.299,temp=303):
'''
input:
pct_us: float, weight percent UO2-SO4 in water
temp: float, temperature in K
output:
rho_us: float, mass density (g/cc) of the solution
This is derived from Table 5 of IEU-SOL-THERM-004
'''
temp = temp-273.; # convert to C
y = [0.51, 0.399, 0.299];
x = [15.,20.,25.,30.,35.,40.,50.];
z = [[1.7959,1.7926,1.7891,1.7854,1.7816,1.777,1.7696],
[1.5283,1.5257,1.5228,1.5199,1.5166,1.5133,1.5064],
[1.3499,1.3479,1.3455,1.3430,1.3403,1.3376,1.3314]];
f = scipy.interpolate.interp2d(x,y,z);
rho_us = np.float64(f(temp,pct_us));
return rho_us
# the atomic weight of enriched uranium is needed
def get_Aw_U(enrichment):
assert (enrichment<=1.0), "enrichment should be entered as a percentage"
# assumes a fixed ratio in the percentage of U234 and U235.
# i.e. the enrichment process keeps these two isotopes in the
# same relative abundance
U234_to_U235_ratio = 0.0055/0.72;
Aw_U235 = openmc.data.atomic_mass('U235');
Aw_U234 = openmc.data.atomic_mass('U234');
Aw_U238 = openmc.data.atomic_mass('U238');
frac_235 = enrichment;
frac_234 = frac_235*U234_to_U235_ratio;
frac_238 = 1. - frac_235 - frac_234;
aw = 1./(frac_235/Aw_U235 + frac_238/Aw_U238 + frac_234/Aw_U234);
weight_frac = {};
weight_frac['U234']=frac_234;
weight_frac['U235']=frac_235;
weight_frac['U238']=frac_238;
return aw,weight_frac
# calculate atom densities of UO2-SO4 + H2O solution as a function
# of Uranium enrichment, water temperature and Urynal Sulfate concentration
def BoilerAtomDensities(enrich=0.1467,temp=303,conc=0.299):
'''
input:
enrich: w/o U-235
temp: solution temperature in K;
conc: w/o concentration of Uranyl Sulfate in the water
output:
dictionary with the atom densities (atoms/b-cm) of all elements
and nuclides in the solution
this results in a 'not great, not terrible' agreement with the benchmark.
must re-visit to make corrections
'''
assert (temp > 288) and (temp < 323), 'temperature not in correlation limits';
assert (conc >= 0.299) and (conc <= 0.51), 'solution concentration not in correlation limits';
Na = 0.60221; # Avagadro's number x10**-24
AtomDensities = {};
rho_uranyl_sulf_sol = get_UranylSulfate_solution_density(conc,temp);
rho_water = rho_uranyl_sulf_sol*(1-conc);
rho_uranyl = rho_uranyl_sulf_sol*conc;
Aw_234 = openmc.data.atomic_mass('U234');
Aw_235 = openmc.data.atomic_mass('U235');
Aw_238 = openmc.data.atomic_mass('U238');
Aw_S = openmc.data.atomic_weight('S');
Aw_H = openmc.data.atomic_weight('H');
Aw_O = openmc.data.atomic_weight('O');
Aw_U,U_weight_fracs = get_Aw_U(enrich);
Aw_uranyl_sulf = Aw_U + 2.*Aw_O + Aw_S + 4.*Aw_O;
Aw_h2o = 2.*Aw_H + Aw_O;
mol_density_uranyl_sulf = (rho_uranyl/Aw_uranyl_sulf)*Na;
mol_density_h2o = (rho_water/Aw_h2o)*Na;
AtomDensities['H']=mol_density_h2o*2.;
AtomDensities['O']=mol_density_h2o*1.;
AtomDensities['O']+=mol_density_uranyl_sulf*4.;
AtomDensities['S']=mol_density_uranyl_sulf*1.;
AtomDensities['U234']=(mol_density_uranyl_sulf)*U_weight_fracs['U234']; #wrong but better
AtomDensities['U235']=(mol_density_uranyl_sulf)*U_weight_fracs['U235'];
AtomDensities['U238']=(mol_density_uranyl_sulf)*U_weight_fracs['U238'];
return AtomDensities
###Output
_____no_output_____
###Markdown
Solution Material
###Code
sol_temp = 303; # K, Uranyl Sulfate and water solution temperature
sol_conc = 0.299; # w/o of Uranyl Sulfate in the solution
U_enrch = 0.1467; # w/o enrichment of U235 in Uranyl Sulfate
sol_atom_densities = BoilerAtomDensities(enrich=U_enrch,temp=sol_temp,conc=sol_conc);
sol = openmc.Material(name='sol');
sol.add_element('H',sol_atom_densities['H'],percent_type='ao');
sol.add_element('O',sol_atom_densities['O'],percent_type='ao');
sol.add_element('S',sol_atom_densities['S'],percent_type='ao');
sol.add_nuclide('U234',sol_atom_densities['U234'],percent_type='ao');
sol.add_nuclide('U235',sol_atom_densities['U235'],percent_type='ao');
sol.add_nuclide('U238',sol_atom_densities['U238'],percent_type='ao');
sol.add_s_alpha_beta('c_H_in_H2O');
ad_tot = 0.;
for key in sol_atom_densities:
ad_tot+=sol_atom_densities[key];
sol.set_density('atom/b-cm',ad_tot);
###Output
_____no_output_____
###Markdown
Shell MaterialI will just use the nuclide and density information from PNNL-15870 Rev. 1 for Steel, Stainless 347. The composition and atom density for the nominal 347 SS material reported in the benchmark closely match the PNNL data.
###Code
shell = openmc.Material(name='shell');
shell.add_element('C',0.003659);
shell.add_element('Si',0.019559);
shell.add_element('P',0.000798);
shell.add_element('S',0.000514);
shell.add_element('Cr',0.179602);
shell.add_element('Mn',0.019998);
shell.add_element('Fe',0.669338);
shell.add_element('Ni',0.102952);
shell.add_element('Nb',0.002365);
shell.add_element('Ta',0.001214);
shell.set_density('g/cc',8.0);
###Output
_____no_output_____
###Markdown
Beryllium Oxide ReflectorThe reported composition and density for the Beryllium Oxide blocks in the benchmark differ *significantly* from PNNL-15870. This is owing to the quick-and-dirty fabrication of the blocks (and associated reduction in material density) as well as the high impurity content.
###Code
beryl_ref = openmc.Material(name='beryl_ref');
beryl_ref.add_element('O',6.6210e-2);
beryl_ref.add_element('Be',6.6210e-2);
beryl_ref.add_element('B',3.0637e-7);
beryl_ref.add_element('Co',5.6202e-7);
beryl_ref.add_element('Ag',3.0706e-8);
beryl_ref.add_element('Cd',7.3662e-8);
beryl_ref.add_element('In',1.4423e-8);
beryl_ref.add_s_alpha_beta('c_Be_in_BeO');
beryl_ref.set_density('g/cc',2.75);
###Output
_____no_output_____
###Markdown
Graphite BaseThe beryllium semi-sphere sits in a base of graphite. No information is given on it's composition. It probably has a trivial impact on the model result but I will just use standard data for graphite
###Code
grph = openmc.Material(name='grph');
grph.add_element('C',0.999999);
grph.add_element('B',0.000001);
grph.set_density('g/cc',1.7);
grph.add_s_alpha_beta('c_Graphite');
###Output
_____no_output_____
###Markdown
Air outside the reactorI don't really care about the air for its interactions with neutrons; I really just want to include it as a material so I can more fully visualize the geometry.
###Code
air = openmc.Material(name='air');
air.add_element('C',0.000150);
air.add_element('N',0.784431);
air.add_element('O',0.210748);
air.add_element('Ar',0.004671);
materials = openmc.Materials([sol,shell,beryl_ref,grph,air]);
materials.export_to_xml();
###Output
_____no_output_____
###Markdown
GeometryCreate the geometry and plot. Not too complicated. Surfaces
###Code
rx_origin = [0.,76.3214,0.];
ref_sphere = openmc.Sphere(y0=rx_origin[1],r=47.4210);
tank_o = openmc.Sphere(y0=rx_origin[1],r=15.3614);
tank_i = openmc.Sphere(y0=rx_origin[1],r=15.282);
graph_base_cyl = openmc.YCylinder(r=47.4210);
fill_drain_cav = openmc.YCylinder(r=4.445/2.);
fill_drain_o = openmc.YCylinder(r=2.06375);
fill_drain_i = openmc.YCylinder(r=1.905);
plate_plane = openmc.YPlane(y0=0.);
base_plane = openmc.YPlane(y0=34.4114);
sphere_center_plane = openmc.YPlane(y0=rx_origin[1]);
upper_plane = openmc.YPlane(y0=118.2314);
bbox = openmc.model.RightCircularCylinder([0.,-10.,0.],140.,60.,axis='y',boundary_type='vacuum');
###Output
_____no_output_____
###Markdown
Cells
###Code
colors = {}
colors[grph]='black';
colors[shell]='silver';
colors[sol]='cadetblue';
colors[beryl_ref]='olive';
colors[air]='lightskyblue';
core = openmc.Cell();
core.fill = sol;
core.region = (-tank_i) | (-fill_drain_i) & -bbox
root = openmc.Universe();
root.add_cell(core);
root.plot(origin=rx_origin,width=[200., 200.],pixels=[800,800],
color_by='material',colors=colors);
steel_tank_and_pipe = openmc.Cell();
steel_tank_and_pipe.fill = shell;
#steel_tank_and_pipe.region = (+tank_i & -tank_o & ~(-fill_drain_i)) | \
# (+fill_drain_i & -fill_drain_o & +tank_i) & -bbox
steel_tank_and_pipe.region = (+tank_i & -tank_o & ~(-fill_drain_i)) | \
(+fill_drain_i & -fill_drain_o & +tank_i) & -bbox;
root = openmc.Universe();
root.add_cell(steel_tank_and_pipe);
root.plot(origin=[0.,76.3214+15.,0.],width=[10., 10.],pixels=[400,400],
color_by='material',colors=colors);
# moved the origin around so the plot can resolve all surface areas in question
root = openmc.Universe();
root.add_cells([core,steel_tank_and_pipe]);
root.plot(origin=[0.,76.3214+15.,0.],width=[10., 10.],pixels=[400,400],
color_by='material',colors=colors);
# moved the origin around so the plot can resolve all surface areas in question
ref = openmc.Cell();
ref.fill = beryl_ref;
ref.region = (+tank_o & +fill_drain_o) & -ref_sphere & +base_plane & -upper_plane;
root = openmc.Universe();
root.add_cells([ref,core,steel_tank_and_pipe]);
root.plot(origin=rx_origin,width=[100., 100.],pixels=[400,400],
color_by='material',colors=colors);
graph_base = openmc.Cell();
graph_base.fill = grph;
graph_base.region = ((-graph_base_cyl & +plate_plane & -base_plane & +fill_drain_o) |
(-graph_base_cyl & +ref_sphere & +base_plane & -sphere_center_plane));
root = openmc.Universe();
root.add_cells([graph_base,ref,core,steel_tank_and_pipe]);
root.plot(origin=rx_origin,width=[150., 180.],pixels=[800,800],
color_by='material',colors=colors);
outside = openmc.Cell();
outside.fill = air;
outside.region = -bbox & (+graph_base_cyl | (+ref_sphere & -upper_plane) |
(+upper_plane & +fill_drain_o) )
root = openmc.Universe();
root.add_cells([graph_base,ref,core,steel_tank_and_pipe,outside]);
root.plot(origin=rx_origin,width=[150., 200.],pixels=[800,800],
color_by='material',colors=colors);
geometry = openmc.Geometry();
geometry.root_universe = root;
geometry.export_to_xml();
###Output
_____no_output_____
###Markdown
Tallies
###Code
cell_filter = openmc.CellFilter(core);
N = 1001;
energy_bins = np.logspace(-3,7,num=N);
energy_filter = openmc.EnergyFilter(values=energy_bins);
abs_core = openmc.Tally(name='abs_core');
abs_core.scores = ['absorption'];
abs_core.filters = [cell_filter,energy_filter];
fission = openmc.Tally(name='fission');
fission.scores = ['fission'];
fission.filters = [cell_filter,energy_filter];
fission_by_nuclide = openmc.Tally(name='fission_by_nuclide');
fission_by_nuclide.scores = ['fission'];
fission_by_nuclide.nuclides = ['U234','U235','U238'];
fission_by_nuclide.filters = [cell_filter,energy_filter];
capture = openmc.Tally(name='capture');
capture.scores = ['(n,gamma)'];
capture.filters = [cell_filter,energy_filter];
capture_by_nuclide = openmc.Tally(name='capture_by_nuclide');
capture_by_nuclide.scores = ['(n,gamma)'];
capture_by_nuclide.nuclides = ['U234','U238','H1','O16','S32'];
capture_by_nuclide.filters = [cell_filter,energy_filter];
flux = openmc.Tally(name='flux');
flux.scores = ['flux'];
flux.filters = [cell_filter,energy_filter];
tallies = openmc.Tallies([abs_core, flux,
fission, capture,
fission_by_nuclide,
capture_by_nuclide]);
tallies.export_to_xml();
settings = openmc.Settings();
settings.batches = 800;
settings.inactive = 100;
settings.particles = 25000;
R = 15.;
y_org = 76.3214;
bounds = [-R,-R+y_org,-R,R,R+y_org,R];
uniform_dist = openmc.stats.Box(bounds[:3],bounds[3:],
only_fissionable=True);
settings.source = openmc.source.Source(space=uniform_dist);
#settings.temperature['method']='interpolation';
settings.export_to_xml();
openmc.run()
sp = openmc.StatePoint('statepoint.800.h5');
sp.tallies
flux = sp.get_tally(name='flux');
flux_df = flux.get_pandas_dataframe();
flux_vals = flux_df['mean'].to_numpy();
energy_x = 0.5*(energy_bins[0:-1] + energy_bins[1:]);
plt.loglog(energy_x,flux_vals);
plt.grid();
plt.xlabel('Energy [eV]');
plt.ylabel('flux [n/cm**2-s]');
###Output
_____no_output_____
###Markdown
Notice the sharp dip in flux at about 10-20 eV. Presumably this is a resonance absorption peak for something in the core; I'd guess U-238. The logical nuclide to suspect is U-238 but there is no good reason not to at least look at the other cross sections, so I will import and plot the capture cross sections for all of the isotopes.
###Code
OMC_DATA = "/home/sblair/OMC_DATA/endfb71_hdf5"
u238_path = OMC_DATA + "/U238.h5";
u238 = openmc.data.IncidentNeutron.from_hdf5(u238_path);
u238_capture = u238[102];
s32_path = OMC_DATA + "/S32.h5";
s32 = openmc.data.IncidentNeutron.from_hdf5(s32_path);
s32_capture = s32[102];
plt.rcParams['figure.figsize']=[12,8];
plt.loglog(energy_x,flux_vals,label='flux');
plt.loglog(u238_capture.xs['294K'].x,u238_capture.xs['294K'].y,label='U238');
plt.loglog(s32_capture.xs['294K'].x,s32_capture.xs['294K'].y,label='S32');
plt.grid();
plt.legend();
plt.xlabel('Energy [eV]');
###Output
_____no_output_____
###Markdown
Notice the alignment of the first 3 large resonance peaks of U-238 with the first 3 prominent "dips" in flux. I plotted S-32 just for reference; I wanted to see the energy at which its resonance region starts. In general, the resonance region for light isotopes begins at a higher energy than the resonance region for heavy isotopes. S-32 is the only other isotope besides the Uranium isotopes that has a reasonably high mass number and non-trivial atom density. We included U-234 in the model and that must be for a reason. Let's plot U-234 capture cross section along with U-238 and the flux to see if that might be a contributor.
###Code
u234_path = OMC_DATA + "/U234.h5";
u234 = openmc.data.IncidentNeutron.from_hdf5(u234_path);
u234_capture = u234[102];
plt.rcParams['figure.figsize']=[12,8];
plt.loglog(energy_x,flux_vals,label='flux');
plt.loglog(u238_capture.xs['294K'].x,u238_capture.xs['294K'].y,label='U-238');
plt.loglog(u234_capture.xs['294K'].x,u234_capture.xs['294K'].y,label='U-234');
plt.grid();
plt.legend();
plt.xlabel('Energy [eV]');
###Output
_____no_output_____
###Markdown
Ahh! So it turns out that U-234 has an even bigger and lower-lying capture resonance than U-238. Without a doubt this is why neutron transport models include this (small) isotope.
###Code
capture_by_nuclide = sp.get_tally(name='capture_by_nuclide');
capture_by_nuclide_df = capture_by_nuclide.get_pandas_dataframe();
capture_U234 = capture_by_nuclide_df[capture_by_nuclide_df['nuclide']=='U234']['mean'].to_numpy();
capture_U238 = capture_by_nuclide_df[capture_by_nuclide_df['nuclide']=='U238']['mean'].to_numpy();
capture_H1 = capture_by_nuclide_df[capture_by_nuclide_df['nuclide']=='H1']['mean'].to_numpy();
capture_S32 = capture_by_nuclide_df[capture_by_nuclide_df['nuclide']=='S32']['mean'].to_numpy();
capture_O16 = capture_by_nuclide_df[capture_by_nuclide_df['nuclide']=='O16']['mean'].to_numpy();
plt.rcParams['figure.figsize']=[12,8];
plt.loglog(energy_x,flux_vals,label='flux');
plt.loglog(energy_x,capture_U234,label='U-234');
plt.loglog(energy_x,capture_U238,label='U-238');
plt.loglog(energy_x,capture_H1,label='H-1');
plt.loglog(energy_x,capture_S32,label='S-32');
plt.loglog(energy_x,capture_O16,label='O-16');
plt.grid();
plt.legend();
plt.xlabel('Energy [eV]');
###Output
_____no_output_____ |
notebooks/Test20170524.ipynb | ###Markdown
SETUP
###Code
import time
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
%matplotlib inline
###Output
_____no_output_____
###Markdown
Autosipper
###Code
# config directory must have "__init__.py" file
# from the 'config' directory, import the following classes:
from config import Motor, ASI_Controller, Autosipper
from config import utils as ut
autosipper = Autosipper(Motor('config/motor.yaml'), ASI_Controller('config/asi_controller.yaml'))
autosipper.coord_frames
from config import gui
gui.stage_control(autosipper.XY, autosipper.Z)
# add/determine deck info
autosipper.coord_frames.deck.position_table = ut.read_delim_pd('config/position_tables/deck')
# check deck alignment
# CLEAR DECK OF OBSTRUCTIONS!!
autosipper.go_to('deck', ['name'],'align')
# add plate
from config import utils as ut
platemap = ut.generate_position_table((8,8),(9,9),93.5)
platemap[]
ut.lookup(platemap)
###Output
_____no_output_____
###Markdown
Manifold
###Code
from config import Manifold
manifold = Manifold('192.168.1.3', 'config/valvemaps/valvemap.csv', 512)
manifold.valvemap[manifold.valvemap.name>0]
def valve_states():
tmp = []
for i in [2,0,14,8]:
status = 'x'
if manifold.read_valve(i):
status = 'o'
tmp.append([status, manifold.valvemap.name.iloc[i]])
return pd.DataFrame(tmp)
tmp = []
for i in range(16):
status = 'x'
if manifold.read_valve(i):
status = 'o'
name = manifold.valvemap.name.iloc[i]
tmp.append([status, name])
pd.DataFrame(tmp).replace(np.nan, '')
name = 'inlet_in'
v = manifold.valvemap['valve'][manifold.valvemap.name==name]
v=14
manifold.depressurize(v)
manifold.pressurize(v)
manifold.exit()
###Output
_____no_output_____
###Markdown
Micromanager
###Code
# !!!! Also must have MM folder on system PATH
# mm_version = 'C:\Micro-Manager-1.4'
# cfg = 'C:\Micro-Manager-1.4\SetupNumber2_05102016.cfg'
mm_version = 'C:\Program Files\Micro-Manager-2.0beta'
cfg = 'C:\Program Files\Micro-Manager-2.0beta\Setup2_20170413.cfg'
import sys
sys.path.insert(0, mm_version) # make it so python can find MMCorePy
import MMCorePy
from PIL import Image
core = MMCorePy.CMMCore()
core.loadSystemConfiguration(cfg)
core.setProperty("Spectra", "White_Enable", "1")
core.waitForDevice("Spectra")
core.setProperty("Cam Andor_Zyla4.2", "Sensitivity/DynamicRange", "16-bit (low noise & high well capacity)") # NEED TO SET CAMERA TO 16 BIT (ceiling 12 BIT = 4096)
core.setProperty("Spectra", "White_Enable", "0")
###Output
_____no_output_____
###Markdown
Preset: 1_PBP ConfigGroup,Channel,1_PBP,TIFilterBlock1,Label,1-PBPPreset: 2_BF ConfigGroup,Channel,2_BF,TIFilterBlock1,Label,2-BFPreset: 3_DAPI ConfigGroup,Channel,3_DAPI,TIFilterBlock1,Label,3-DAPIPreset: 4_eGFP ConfigGroup,Channel,4_eGFP,TIFilterBlock1,Label,4-GFPPreset: 5_Cy5 ConfigGroup,Channel,5_Cy5,TIFilterBlock1,Label,5-Cy5Preset: 6_AttoPhos ConfigGroup,Channel,6_AttoPhos,TIFilterBlock1,Label,6-AttoPhos TEST 4.5 psi, 25 psi valves
###Code
log = []
autosipper.Z.move(93.5)
manifold.depressurize(2)
manifold.depressurize(0)
log.append([time.ctime(time.time()), 'open inlet_in, inlet_out'])
valve_states()
text = 'fluorescence observed'
log.append([time.ctime(time.time()), text])
text = 'CLOSE inlet_out'
manifold.pressurize(0)
log.append([time.ctime(time.time()), text])
text = 'OPEN chip_in, chip_out'
manifold.depressurize(14)
manifold.depressurize(8)
log.append([time.ctime(time.time()), text])
valve_states()
text = 'fill'
log.append([time.ctime(time.time()), text])
manifold.pressurize(8)
#closed all
autosipper.Z.move(93.5)
manifold.depressurize(2)
manifold.depressurize(0)
log.append([time.ctime(time.time()), 'open inlet_in, inlet_out'])
valve_states()
text = 'fluorescence removed'
log.append([time.ctime(time.time()), text])
text = 'CLOSE inlet_out'
manifold.pressurize(0)
log.append([time.ctime(time.time()), text])
text = 'OPEN chip_in, chip_out'
manifold.depressurize(14)
manifold.depressurize(8)
log.append([time.ctime(time.time()), text])
valve_states()
text = 'flush'
log.append([time.ctime(time.time()), text])
manifold.pressurize(8)
for i in [2,0,14,8]:
manifold.pressurize(i)
###Output
_____no_output_____
###Markdown
ACQUISITION
###Code
log
core.setConfig('Channel','2_BF')
core.setProperty(core.getCameraDevice(), "Exposure", 20)
core.snapImage()
img = core.getImage()
plt.imshow(img,cmap='gray')
image = Image.fromarray(img)
# image.save('TESTIMAGE.tif')
import config.utils as ut
position_list = ut.load_mm_positionlist("C:/Users/fordycelab/Desktop/D1_cjm.pos")
position_list
def acquire():
for i in xrange(len(position_list)):
si = str(i)
x,y = position_list[['x','y']].iloc[i]
core.setXYPosition(x,y)
core.waitForDevice(core.getXYStageDevice())
logadd(core, log, 'moved '+si)
core.snapImage()
# core.waitForDevice(core.getCameraDevice())
logadd(core, log, 'snapped '+si)
img = core.getImage()
logadd(core, log, 'got image '+si)
image = Image.fromarray(img)
image.save('images/images_{}.tif'.format(i))
logadd(core, log, 'saved image '+si)
x,y = position_list[['x','y']].iloc[0]
core.setXYPosition(x,y)
core.waitForDevice(core.getXYStageDevice())
logadd(core, log, 'moved '+ str(0))
def logadd(core,log,st):
log.append([time.ctime(time.time()), st])
core.logMessage(st)
print log[-1]
# Trial 1: returning stage to home at end of acquire
log = []
for i in xrange(15):
sleep = (10*i)*60
logadd(log, 'STRT SLEEP '+ str(sleep/60) + ' min')
time.sleep(sleep)
logadd(log, 'ACQ STARTED '+str(i))
acquire()
core.stopSecondaryLogFile(l2)
# Trial 2: returning stage to home at end of acquire
# added mm logs
# core.setPrimaryLogFile('20170524_log_prim2.txt')
# core.enableDebugLog(True)
# core.enableStderrLog(True)
l2 = core.startSecondaryLogFile('20170524_log_sec3.txt', True, False, True)
log = []
for i in xrange(15):
sleep = (10*i)*60
logadd(core, log, 'STRT SLEEP '+ str(sleep/60) + ' min')
time.sleep(sleep)
logadd(core, log, 'ACQ STARTED '+str(i))
acquire()
core.stopSecondaryLogFile(l2)
# Trial 3: returning stage to home at end of acquire
# added mm logs
# core.setPrimaryLogFile('20170524_log_prim2.txt')
# core.enableDebugLog(True)
# core.enableStderrLog(True)
l2 = core.startSecondaryLogFile('20170524_log_sec3.txt', True, False, True)
log = []
for i in xrange(15):
sleep = (10*i)*60
logadd(core, log, 'STRT SLEEP '+ str(sleep/60) + ' min')
time.sleep(sleep)
logadd(core, log, 'ACQ STARTED '+str(i))
acquire()
core.stopSecondaryLogFile(l2)
# Trial 4: returning stage to home at end of acquire
# added mm logs
# after USB fix
# core.setPrimaryLogFile('20170524_log_prim2.txt')
# core.enableDebugLog(True)
# core.enableStderrLog(True)
l2 = core.startSecondaryLogFile('20170526_log_sec4.txt', True, False, True)
log = []
for i in xrange(15):
sleep = (5*i)*60
logadd(core, log, 'STRT SLEEP '+ str(sleep/60) + ' min')
time.sleep(sleep)
logadd(core, log, 'ACQ STARTED '+str(i))
acquire()
core.stopSecondaryLogFile(l2)
# Trial 5: returning stage to home at end of acquire
# second one after USB fix
# core.setPrimaryLogFile('20170524_log_prim2.txt')
# core.enableDebugLog(True)
# core.enableStderrLog(True)
l2 = core.startSecondaryLogFile('20170526_log_sec5.txt', True, False, True)
log = []
for i in xrange(15):
sleep = (10*i)*60
logadd(core, log, 'STRT SLEEP '+ str(sleep/60) + ' min')
time.sleep(sleep)
logadd(core, log, 'ACQ STARTED '+str(i))
acquire()
core.stopSecondaryLogFile(l2)
# Auto
core.setAutoShutter(True) # default
core.snapImage()
# Manual
core.setAutoShutter(False) # disable auto shutter
core.setProperty("Shutter", "State", "1")
core.waitForDevice("Shutter")
core.snapImage()
core.setProperty("Shutter", "State", "0")
###Output
_____no_output_____
###Markdown
MM Get info
###Code
core.getFocusDevice()
core.getCameraDevice()
core.XYStageDevice()
core.getDevicePropertyNames(core.getCameraDevice())
###Output
_____no_output_____
###Markdown
Video
###Code
import cv2
from IPython import display
import numpy as np
from ipywidgets import widgets
import time
# core.initializeCircularBuffer()
# core.setCircularBufferMemoryFootprint(4096) # MiB
cv2.WND
# video with button (CV2)
live = widgets.Button(description='Live')
close = widgets.Button(description='Close')
display.display(widgets.HBox([live, close]))
def on_live_clicked(b):
display.clear_output(wait=True)
print 'LIVE'
core.startContinuousSequenceAcquisition(1000) # time overridden by exposure
time.sleep(.2)
cv2.namedWindow('Video', cv2.WINDOW_NORMAL)
cv2.setWindowProperty('Video', cv2.WND_PROP_ASPECT_RATIO, cv2.WINDOW_KEEPRATIO)
cv2.resizeWindow('Video', 500,500)
img = np.zeros((500,500))
print 'To stop, click window + press ESC'
while(1):
time.sleep(.015)
if core.getRemainingImageCount() > 0:
img = core.getLastImage()
cv2.imshow('Video',img)
k = cv2.waitKey(30)
if k==27: # ESC key; may need 255 mask?
break
print 'STOPPED'
core.stopSequenceAcquisition()
def on_close_clicked(b):
if core.isSequenceRunning():
core.stopSequenceAcquisition()
cv2.destroyWindow('Video')
live.on_click(on_live_clicked)
close.on_click(on_close_clicked)
# video with button (CV2)
# serial snap image
live = widgets.Button(description='Live')
close = widgets.Button(description='Close')
display.display(widgets.HBox([live, close]))
def on_live_clicked(b):
display.clear_output(wait=True)
print 'LIVE'
cv2.namedWindow('Video', cv2.WINDOW_NORMAL)
cv2.setWindowProperty('Video', cv2.WND_PROP_ASPECT_RATIO, cv2.WINDOW_KEEPRATIO)
cv2.resizeWindow('Video', 500,500)
img = np.zeros((500,500))
print 'To stop, click window + press ESC'
while(1):
core.snapImage()
time.sleep(.05)
img = core.getImage()
cv2.imshow('Video',img)
k = cv2.waitKey(30)
if k==27: # ESC key; may need 255 mask?
break
print 'STOPPED'
def on_close_clicked(b):
if core.isSequenceRunning():
core.stopSequenceAcquisition()
cv2.destroyWindow('Video')
live.on_click(on_live_clicked)
close.on_click(on_close_clicked)
cv2.destroyAllWindows()
###Output
_____no_output_____
###Markdown
SNAP CV2
###Code
# snap (CV2)
snap = widgets.Button(description='Snap')
close2 = widgets.Button(description='Close')
display.display(widgets.HBox([snap, close2]))
def on_snap_clicked(b):
cv2.destroyWindow('Snap')
cv2.namedWindow('Snap',cv2.WINDOW_NORMAL)
cv2.resizeWindow('Snap', 500,500)
cv2.setWindowProperty('Snap', cv2.WND_PROP_ASPECT_RATIO, cv2.WINDOW_KEEPRATIO)
core.snapImage()
time.sleep(.1)
img = core.getImage()
cv2.imshow('Snap',img)
k = cv2.waitKey(30)
def on_close2_clicked(b):
cv2.destroyWindow('Snap')
snap.on_click(on_snap_clicked)
close2.on_click(on_close2_clicked)
###Output
_____no_output_____
###Markdown
EXIT
###Code
autosipper.exit()
manifold.exit()
core.unloadAllDevices()
core.reset()
print 'closed'
import config.gui as gui
core.
gui.video(core)
gui.snap(core, 'mpl')
gui.manifold_control(manifold)
###Output
_____no_output_____ |
Paper_code/Ensemble_FullData.ipynb | ###Markdown
Ensemble decoder on full dataset User Options
###Code
# dataset='s1'
# dataset='m1'
dataset='hc'
#What folder to save the ensemble results to
#Note that the data we are loading are in this same folder (since they are the results from the other decoders)
save_folder=''
# save_folder='/home/jglaser/Files/Neural_Decoding/Results/'
###Output
_____no_output_____
###Markdown
Import packages
###Code
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline
from scipy import io
from scipy import stats
import pickle
import sys
#Add the main folder to the path, so we have access to the files there.
#Note that if your working directory is not the Paper_code folder, you may need to manually specify the path to the main folder. For example: sys.path.append('/home/jglaser/GitProj/Neural_Decoding')
sys.path.append('..')
#Import metrics
from metrics import get_R2
from decoders import DenseNNDecoder
from bayes_opt import BayesianOptimization
#Turn off deprecation warnings
import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning)
###Output
_____no_output_____
###Markdown
Load results from other decodersNote we do not use the Kalman filter results in our ensemble due to slightly different formatting
###Code
with open(save_folder+dataset+'_ground_truth.pickle','rb') as f:
[y_test_all,y_train_all,y_valid_all]=pickle.load(f)
with open(save_folder+dataset+'_results_wf2.pickle','rb') as f:
[mean_r2_wf,y_pred_wf_all,y_train_pred_wf_all,y_valid_pred_wf_all]=pickle.load(f)
with open(save_folder+dataset+'_results_wc2.pickle','rb') as f:
[mean_r2_wc,y_pred_wc_all,y_train_pred_wc_all,y_valid_pred_wc_all]=pickle.load(f)
with open(save_folder+dataset+'_results_xgb2.pickle','rb') as f:
[mean_r2_xgb,y_pred_xgb_all,y_train_pred_xgb_all,y_valid_pred_xgb_all,time_elapsed]=pickle.load(f)
with open(save_folder+dataset+'_results_svr2.pickle','rb') as f:
[mean_r2_svr,y_pred_svr_all,y_train_pred_svr_all,y_valid_pred_svr_all,time_elapsed]=pickle.load(f)
with open(save_folder+dataset+'_results_dnn2.pickle','rb') as f:
[mean_r2_dnn,y_pred_dnn_all,y_train_pred_dnn_all,y_valid_pred_dnn_all,time_elapsed]=pickle.load(f)
with open(save_folder+dataset+'_results_rnn2.pickle','rb') as f:
[mean_r2_rnn,y_pred_rnn_all,y_train_pred_rnn_all,y_valid_pred_rnn_all,time_elapsed]=pickle.load(f)
with open(save_folder+dataset+'_results_gru2.pickle','rb') as f:
[mean_r2_gru,y_pred_gru_all,y_train_pred_gru_all,y_valid_pred_gru_all,time_elapsed]=pickle.load(f)
with open(save_folder+dataset+'_results_lstm2.pickle','rb') as f:
[mean_r2_lstm,y_pred_lstm_all,y_train_pred_lstm_all,y_valid_pred_lstm_all,time_elapsed]=pickle.load(f)
###Output
_____no_output_____
###Markdown
Run ensemble method1. We loop through each CV fold and both out (x and y position/velocities).2. We create the matrix of covariates (the predictions from the other methods)3. We optimize the hyperparameters for the fully connected (dense) neural network we are using, based on validation set R2 values4. We fit the neural net on training data w/ the optimal hyperparameters5. We make test set predictions and get test set R2 values
###Code
##Initialize
y_pred_ensemble_all=[] #List where test set predictions are put (for saving and plotting)
mean_r2_dnn=np.empty([10,2]) #Where the R2 values are saved (matrix of 10 CV folds x 2 outputs)
for i in range(10): #Loop through the cross validation folds
for j in range(2): #Loop through the 2 output predictions (x and y positions/velocities)
###CREATE COVARIATES###
#Make matrix of covariates, where each feature is the predictions from one of the other decoders
#Do this for training, validation, and testing data
X_train=np.concatenate((y_train_pred_wf_all[i][:,j:j+1], y_train_pred_wc_all[i][:,j:j+1],
y_train_pred_svr_all[i][:,j:j+1],y_train_pred_xgb_all[i][:,j:j+1],
y_train_pred_dnn_all[i][:,j:j+1], y_train_pred_rnn_all[i][:,j:j+1],
y_train_pred_gru_all[i][:,j:j+1], y_train_pred_lstm_all[i][:,j:j+1]),axis=1)
X_valid=np.concatenate((y_valid_pred_wf_all[i][:,j:j+1], y_valid_pred_wc_all[i][:,j:j+1],
y_valid_pred_svr_all[i][:,j:j+1],y_valid_pred_xgb_all[i][:,j:j+1],
y_valid_pred_dnn_all[i][:,j:j+1], y_valid_pred_rnn_all[i][:,j:j+1],
y_valid_pred_gru_all[i][:,j:j+1], y_valid_pred_lstm_all[i][:,j:j+1]),axis=1)
X_test=np.concatenate((y_pred_wf_all[i][:,j:j+1], y_pred_wc_all[i][:,j:j+1],
y_pred_svr_all[i][:,j:j+1],y_pred_xgb_all[i][:,j:j+1],
y_pred_dnn_all[i][:,j:j+1], y_pred_rnn_all[i][:,j:j+1],
y_pred_gru_all[i][:,j:j+1], y_pred_lstm_all[i][:,j:j+1]),axis=1)
#Get outputs (training/validation/testing) for this CV fold and output
y_train=y_train_all[i][:,j:j+1]
y_valid=y_valid_all[i][:,j:j+1]
y_test=y_test_all[i][:,j:j+1]
###HYPERPARAMETER OPTIMIZATION###
#Define a function that returns the metric we are trying to optimize (R2 value of the validation set)
#as a function of the hyperparameter we are fitting (num_units, frac_dropout, n_epochs)
def dnn_evaluate(num_units,frac_dropout,n_epochs):
num_units=int(num_units) #Put in proper format (Bayesian optimization uses floats, and we just want to test the integer)
frac_dropout=float(frac_dropout) #Put in proper format
n_epochs=int(n_epochs) #Put in proper format
model_dnn=DenseNNDecoder(units=[num_units,num_units],dropout=frac_dropout,num_epochs=n_epochs) #Define model
model_dnn.fit(X_train,y_train) #Fit model
y_valid_predicted_dnn=model_dnn.predict(X_valid) #Get validation set predictions
return np.mean(get_R2(y_valid,y_valid_predicted_dnn)) #Return mean validation set R2
#Do bayesian optimization
dnnBO = BayesianOptimization(dnn_evaluate, {'num_units': (3, 50), 'frac_dropout': (0,.5), 'n_epochs': (2,10)}, verbose=0) #Define Bayesian optimization, and set limits of hyperparameters
#Set number of initial runs and subsequent tests, and do the optimization. Also, we set kappa=10 (greater than the default) so there is more exploration when there are more hyperparameters
dnnBO.maximize(init_points=10, n_iter=15, kappa=10)
best_params=dnnBO.res['max']['max_params'] #Get the hyperparameters that give rise to the best fit
num_units=np.int(best_params['num_units']) #We want the integer value associated with the best "num_units" parameter (which is what the xgb_evaluate function does above)
frac_dropout=float(best_params['frac_dropout'])
n_epochs=np.int(best_params['n_epochs'])
# Run model w/ above hyperparameters
model_dnn=DenseNNDecoder(units=[num_units,num_units],dropout=frac_dropout,num_epochs=n_epochs) #Declare model w/ fit hyperparameters
model_dnn.fit(X_train,y_train) #Fit model
y_test_predicted_dnn=model_dnn.predict(X_test) #Get test set predictions
mean_r2_dnn[i,j]=np.mean(get_R2(y_test,y_test_predicted_dnn)) #Get test set R2
#Print R2 values
R2s_dnn=get_R2(y_test,y_test_predicted_dnn)
print('R2s:', R2s_dnn)
y_pred_ensemble_all.append(y_test_predicted_dnn) # #Add test set predictions to list (for saving)
mean_r2_ensemble=np.mean(mean_r2_dnn,axis=1) #Get mean R2 value for each fold (across x and y predictions)
#Save data
with open(save_folder+dataset+'_results_ensemble_dnn2.pickle','wb') as f:
pickle.dump([mean_r2_ensemble,y_pred_ensemble_all],f)
###Output
/opt/anaconda/anaconda2/lib/python2.7/site-packages/keras/models.py:826: UserWarning: The `nb_epoch` argument in `fit` has been renamed `epochs`.
warnings.warn('The `nb_epoch` argument in `fit` '
|
src/jupyterNotebook.ipynb | ###Markdown
Welcome to a simple jupyter lab tutorial fileThis cell is rendered in markdown
###Code
# This cell is in python code
###Output
_____no_output_____
###Markdown
Interactive EnvironmentJupyter Lab is great for development and data science because output appears immediately below the code block:
###Code
# lets output some random numbers!
import numpy as np
rng = np.random.default_rng(seed=42)
array = rng.random((100, 1))
array[:20]
###Output
_____no_output_____
###Markdown
PlottingWe can also display plots in the notebook without calling external windows/plotters
###Code
import matplotlib.pyplot as plt
# data to be plotted
x = np.arange(1, 101)
y = array
# plotting
plt.title("Random Sampling")
plt.xlabel("sample")
plt.ylabel("value")
plt.plot(x, y, color ="red")
plt.show()
###Output
_____no_output_____ |
week2/vishwajeet/Q3 - Q/Attempt1_filesubmission_Vishwajeet_search_based_planning_package.ipynb | ###Markdown
We are going to use networkx package to construct the graph and find the shortest paths. Refer to the [NetworkX documentation](https://networkx.github.io/documentation/stable/).
###Code
#type in the edges and edgecost as a list of 3-tuples
edges = [(0,1,2),(0,2, 1.5),(0,3, 2.5),(1,4, 1.5),(2,5, 0.5),(4,8, 1),
(2,6, 2.5),(3,7, 2),(7,9, 1.25),(5,10, 2.75),(6,10, 3.25),
(9,10, 1.5),(8,10, 3.5)]
#Define an empty graph
G = nx.Graph()
#populate the edges and the cost in graph G
G.add_weighted_edges_from(edges,weight='cost')
list(G.nodes)
#Find the shortest path from Node 0 to Node 10
print(nx.shortest_path(G,0,10,'cost'))
#Find the cost of the shortest path from Node 0 to Node 10
print(nx.shortest_path_length(G,0,10,'cost'))
###Output
[0, 2, 5, 10]
4.75
###Markdown
Let us now move onto a grid which represents the robot's operating environment. First convert the grid to a graph. Then we will use Astar from networkX to find the shortest path
###Code
# write the Euclidean function that takes in the
# node x, y and compute the distance
def euclidean(node1, node2):
x1, y1 = node1
x2, y2 = node2
return np.sqrt((x1-x2)**2+(y1-y2)**2)
# use np.load to load a grid of 1s and 0s
# 1 - occupied 0- free
grid = np.load("astar_grid.npy")
print(grid)
# you can define your own start/ end
start = (0, 0)
goal = (0, 19)
# visualize the start/ end and the robot's environment
fig, ax = plt.subplots(figsize=(12,12))
ax.imshow(grid, cmap=plt.cm.Dark2)
ax.scatter(start[1],start[0], marker = "+", color = "yellow", s = 200)
ax.scatter(goal[1],goal[0], marker = "+", color = "red", s = 200)
plt.show()
###Output
_____no_output_____
###Markdown
Convert this grid array into a graph. You have to follow these steps1. Find the dimensions of grid. Use grid_2d_graph() to initialize a grid graph of corresponding dimensions2. Use remove_node() to remove nodes and edges of all cells that are occupied
###Code
#initialize graph
grid_size=grid.shape
G=nx.grid_2d_graph(*grid_size)
deleted_nodes = 0 # counter to keep track of deleted nodes
#loop to remove nodes
for i in range(grid_size[0]):
for j in range(grid_size[1]):
if grid[i,j]==1:
G.remove_node((i,j))
deleted_nodes+=1
print(f"removed {deleted_nodes} nodes")
print(f"number of occupied cells in grid {np.sum(grid)}")
###Output
_____no_output_____
###Markdown
Visualize the resulting graph using nx.draw(). Note that pos argument for nx.draw() has been given below. The graph is too dense. Try changing the node_size and node_color. You can correlate this graph with the grid's occupied cells
###Code
pos = {(x,y):(y,-x) for x,y in G.nodes()}
nx.draw(G, pos=pos, node_color='red', node_size=10)
###Output
_____no_output_____
###Markdown
We are 2 more steps away from finding the path!1. Set edge attribute. Use set_edge_attributes(). Remember we have to provide a dictionary input: Edge is the key and cost is the value. We can set every move to a neighbor to have unit cost.2. Use astar_path() to find the path. Set heuristic to be euclidean distance. weight to be the attribute you assigned in step 1
###Code
nx.set_edge_attributes(G, {e: 1 for e in G.edges()}, "cost")
astar_path = nx.astar_path(G, start, goal, heuristic=euclidean, weight="cost")
astar_path
###Output
_____no_output_____
###Markdown
Visualize the path you have computed!
###Code
fig, ax = plt.subplots(figsize=(12,12))
ax.imshow(grid, cmap=plt.cm.Dark2)
ax.scatter(start[1],start[0], marker = "+", color = "yellow", s = 200)
ax.scatter(goal[1],goal[0], marker = "+", color = "red", s = 200)
for s in astar_path[1:]:
ax.plot(s[1], s[0],'r+')
###Output
_____no_output_____
###Markdown
Cool! Now you can read arbitrary evironments and find the shortest path between 2 robot positions. Pick a game environment from here and repeat: {https://www.movingai.com/benchmarks/dao/index.html}
###Code
from PIL import Image
import numpy
img= Image.open("combat.png")
np_img = numpy.array(img)
print(np_img.shape)
np_img
###Output
_____no_output_____ |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.