markdown
stringlengths 0
37k
| code
stringlengths 1
33.3k
| path
stringlengths 8
215
| repo_name
stringlengths 6
77
| license
stringclasses 15
values |
|---|---|---|---|---|
Then we import it into a SQLite3 database. The following function automatically guesses the table schema. | from pyensae.sql import import_flatfile_into_database
import_flatfile_into_database("velib_vanves.db3", "velib_vanves.txt", add_key="key") | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
We check the database exists: | import os
os.listdir(".") | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
On Windows, you can use SQLiteSpy to visualize the created table. We use pymysintall to download it. | try:
from pymyinstall.installcustom import install_sqlitespy
exe = install_sqlitespy()
except:
# we skip an exception
# the website can be down...
exe = None
exe | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
We just need to run it (see run_cmd). | if exe:
from pyquickhelper import run_cmd
run_cmd("SQLiteSpy.exe velib_vanves.db3") | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
You should be able to see something like (on Windows): | from pyquickhelper.helpgen import NbImage
NbImage('img_nb_sqlitespy.png') | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
It is easier to use that tool to extract a sample of the data. Once it is ready, you can execute the SQL query in Python and converts the results into a DataFrame. The following code extracts a random sample from the original sets. | sql = """SELECT * FROM velib_vanves WHERE key IN ({0})"""
import random
from pyquickhelper.loghelper import noLOG
from pyensae.sql import Database
db = Database("velib_vanves.db3", LOG = noLOG)
db.connect()
mx = db.execute_view("SELECT MAX(key) FROM velib_vanves")[0][0]
rnd_ids = [ random.randint(1,mx) for i in range(0,100) ] # liste de 100 id aléatoires
strids = ",".join( str(_) for _ in rnd_ids )
res = db.execute_view(sql.format (strids))
df = db.to_df(sql.format (strids))
db.close()
df.head()[["key","last_update","available_bike_stands","available_bikes"]] | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
<h3 id="mem">Memory Dump</h3>
Once you have a big dataset available in text format, it takes some time to load into memory and you need to do that every time you need it again after you closed your python instance. | with open("temp_big_file.txt","w") as f :
f.write("c1\tc2\tc3\n")
for i in range(0,10000000):
x = [ i, random.random(), random.random() ]
s = [ str(_) for _ in x ]
f.write( "\t".join(s) + "\n" )
os.stat("temp_big_file.txt").st_size
import pandas,time
t = time.perf_counter()
df = pandas.read_csv("temp_big_file.txt",sep="\t")
print("duration (s)",time.perf_counter()-t) | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
It is slow considering that many datasets contain many more features. But we can speed it up by doing a kind of memory dump with to_pickle. | t = time.perf_counter()
df.to_pickle("temp_big_file.bin")
print("duration (s)",time.perf_counter()-t) | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
And we reload it with read_pickle: | t = time.perf_counter()
df = pandas.read_pickle("temp_big_file.bin")
print("duration (s)",time.perf_counter()-t) | _doc/notebooks/pyensae_flat2db3.ipynb | sdpython/pyensae | mit |
First fetch from the primary source in s3 as per bug 1312006. We fall back to the github location if this is not available. | import boto3
import botocore
import json
import tempfile
import urllib2
def fetch_schema():
""" Fetch the crash data schema from an s3 location or github location. This
returns the corresponding JSON schema in a python dictionary. """
region = "us-west-2"
bucket = "org-mozilla-telemetry-crashes"
key = "crash_report.json"
fallback_url = "https://raw.githubusercontent.com/mozilla/socorro/master/socorro/schemas/crash_report.json"
try:
log.info("Fetching latest crash data schema from s3://{}/{}".format(bucket, key))
s3 = boto3.client('s3', region_name=region)
# download schema to memory via a file like object
resp = tempfile.TemporaryFile()
s3.download_fileobj(bucket, key, resp)
resp.seek(0)
except botocore.exceptions.ClientError as e:
log.warning(("Could not fetch schema from s3://{}/{}: {}\n"
"Fetching crash data schema from {}")
.format(bucket, key, e, fallback_url))
resp = urllib2.urlopen(fallback_url)
return json.load(resp) | reports/socorro_import/ImportCrashData.ipynb | acmiyaguchi/data-pipeline | mpl-2.0 |
Read crash data as json, convert it to parquet | from datetime import datetime as dt, timedelta, date
from pyspark.sql import SQLContext
def daterange(start_date, end_date):
for n in range(int((end_date - start_date).days) + 1):
yield (end_date - timedelta(n)).strftime("%Y%m%d")
def import_day(d, schema, version):
"""Convert JSON data stored in an S3 bucket into parquet, indexed by crash_date."""
source_s3path = "s3://org-mozilla-telemetry-crashes/v1/crash_report"
dest_s3path = "s3://telemetry-parquet/socorro_crash/"
num_partitions = 10
log.info("Processing {}, started at {}".format(d, dt.utcnow()))
cur_source_s3path = "{}/{}".format(source_s3path, d)
cur_dest_s3path = "{}/v{}/crash_date={}".format(dest_s3path, version, d)
df = sqlContext.read.json(cur_source_s3path, schema=schema)
df.repartition(num_partitions).write.parquet(cur_dest_s3path, mode="overwrite")
def backfill(start_date_yyyymmdd, schema, version):
""" Import data from a start date to yesterday's date.
Example:
backfill("20160902", crash_schema, version)
"""
start_date = dt.strptime(start_date_yyyymmdd, "%Y%m%d")
end_date = dt.utcnow() - timedelta(1) # yesterday
for d in daterange(start_date, end_date):
try:
import_day(d)
except Exception as e:
log.error(e)
from os import environ
# get the relevant date
yesterday = dt.strftime(dt.utcnow() - timedelta(1), "%Y%m%d")
target_date = environ.get('date', yesterday)
# fetch and generate the schema
schema_data = fetch_schema()
crash_schema = create_struct(schema_data)
version = schema_data.get('$target_version', 0) # default to v0
# process the data
import_day(target_date, crash_schema, version) | reports/socorro_import/ImportCrashData.ipynb | acmiyaguchi/data-pipeline | mpl-2.0 |
Downloading the MosMedData: Chest CT Scans with COVID-19 Related Findings
In this example, we use a subset of the
MosMedData: Chest CT Scans with COVID-19 Related Findings.
This dataset consists of lung CT scans with COVID-19 related findings, as well as without such findings.
We will be using the associated radiological findings of the CT scans as labels to build
a classifier to predict presence of viral pneumonia.
Hence, the task is a binary classification problem. | # Download url of normal CT scans.
url = "https://github.com/hasibzunair/3D-image-classification-tutorial/releases/download/v0.2/CT-0.zip"
filename = os.path.join(os.getcwd(), "CT-0.zip")
keras.utils.get_file(filename, url)
# Download url of abnormal CT scans.
url = "https://github.com/hasibzunair/3D-image-classification-tutorial/releases/download/v0.2/CT-23.zip"
filename = os.path.join(os.getcwd(), "CT-23.zip")
keras.utils.get_file(filename, url)
# Make a directory to store the data.
os.makedirs("MosMedData")
# Unzip data in the newly created directory.
with zipfile.ZipFile("CT-0.zip", "r") as z_fp:
z_fp.extractall("./MosMedData/")
with zipfile.ZipFile("CT-23.zip", "r") as z_fp:
z_fp.extractall("./MosMedData/") | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Loading data and preprocessing
The files are provided in Nifti format with the extension .nii. To read the
scans, we use the nibabel package.
You can install the package via pip install nibabel. CT scans store raw voxel
intensity in Hounsfield units (HU). They range from -1024 to above 2000 in this dataset.
Above 400 are bones with different radiointensity, so this is used as a higher bound. A threshold
between -1000 and 400 is commonly used to normalize CT scans.
To process the data, we do the following:
We first rotate the volumes by 90 degrees, so the orientation is fixed
We scale the HU values to be between 0 and 1.
We resize width, height and depth.
Here we define several helper functions to process the data. These functions
will be used when building training and validation datasets. |
import nibabel as nib
from scipy import ndimage
def read_nifti_file(filepath):
"""Read and load volume"""
# Read file
scan = nib.load(filepath)
# Get raw data
scan = scan.get_fdata()
return scan
def normalize(volume):
"""Normalize the volume"""
min = -1000
max = 400
volume[volume < min] = min
volume[volume > max] = max
volume = (volume - min) / (max - min)
volume = volume.astype("float32")
return volume
def resize_volume(img):
"""Resize across z-axis"""
# Set the desired depth
desired_depth = 64
desired_width = 128
desired_height = 128
# Get current depth
current_depth = img.shape[-1]
current_width = img.shape[0]
current_height = img.shape[1]
# Compute depth factor
depth = current_depth / desired_depth
width = current_width / desired_width
height = current_height / desired_height
depth_factor = 1 / depth
width_factor = 1 / width
height_factor = 1 / height
# Rotate
img = ndimage.rotate(img, 90, reshape=False)
# Resize across z-axis
img = ndimage.zoom(img, (width_factor, height_factor, depth_factor), order=1)
return img
def process_scan(path):
"""Read and resize volume"""
# Read scan
volume = read_nifti_file(path)
# Normalize
volume = normalize(volume)
# Resize width, height and depth
volume = resize_volume(volume)
return volume
| examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Let's read the paths of the CT scans from the class directories. | # Folder "CT-0" consist of CT scans having normal lung tissue,
# no CT-signs of viral pneumonia.
normal_scan_paths = [
os.path.join(os.getcwd(), "MosMedData/CT-0", x)
for x in os.listdir("MosMedData/CT-0")
]
# Folder "CT-23" consist of CT scans having several ground-glass opacifications,
# involvement of lung parenchyma.
abnormal_scan_paths = [
os.path.join(os.getcwd(), "MosMedData/CT-23", x)
for x in os.listdir("MosMedData/CT-23")
]
print("CT scans with normal lung tissue: " + str(len(normal_scan_paths)))
print("CT scans with abnormal lung tissue: " + str(len(abnormal_scan_paths)))
| examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Build train and validation datasets
Read the scans from the class directories and assign labels. Downsample the scans to have
shape of 128x128x64. Rescale the raw HU values to the range 0 to 1.
Lastly, split the dataset into train and validation subsets. | # Read and process the scans.
# Each scan is resized across height, width, and depth and rescaled.
abnormal_scans = np.array([process_scan(path) for path in abnormal_scan_paths])
normal_scans = np.array([process_scan(path) for path in normal_scan_paths])
# For the CT scans having presence of viral pneumonia
# assign 1, for the normal ones assign 0.
abnormal_labels = np.array([1 for _ in range(len(abnormal_scans))])
normal_labels = np.array([0 for _ in range(len(normal_scans))])
# Split data in the ratio 70-30 for training and validation.
x_train = np.concatenate((abnormal_scans[:70], normal_scans[:70]), axis=0)
y_train = np.concatenate((abnormal_labels[:70], normal_labels[:70]), axis=0)
x_val = np.concatenate((abnormal_scans[70:], normal_scans[70:]), axis=0)
y_val = np.concatenate((abnormal_labels[70:], normal_labels[70:]), axis=0)
print(
"Number of samples in train and validation are %d and %d."
% (x_train.shape[0], x_val.shape[0])
) | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Data augmentation
The CT scans also augmented by rotating at random angles during training. Since
the data is stored in rank-3 tensors of shape (samples, height, width, depth),
we add a dimension of size 1 at axis 4 to be able to perform 3D convolutions on
the data. The new shape is thus (samples, height, width, depth, 1). There are
different kinds of preprocessing and augmentation techniques out there,
this example shows a few simple ones to get started. | import random
from scipy import ndimage
@tf.function
def rotate(volume):
"""Rotate the volume by a few degrees"""
def scipy_rotate(volume):
# define some rotation angles
angles = [-20, -10, -5, 5, 10, 20]
# pick angles at random
angle = random.choice(angles)
# rotate volume
volume = ndimage.rotate(volume, angle, reshape=False)
volume[volume < 0] = 0
volume[volume > 1] = 1
return volume
augmented_volume = tf.numpy_function(scipy_rotate, [volume], tf.float32)
return augmented_volume
def train_preprocessing(volume, label):
"""Process training data by rotating and adding a channel."""
# Rotate volume
volume = rotate(volume)
volume = tf.expand_dims(volume, axis=3)
return volume, label
def validation_preprocessing(volume, label):
"""Process validation data by only adding a channel."""
volume = tf.expand_dims(volume, axis=3)
return volume, label
| examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
While defining the train and validation data loader, the training data is passed through
and augmentation function which randomly rotates volume at different angles. Note that both
training and validation data are already rescaled to have values between 0 and 1. | # Define data loaders.
train_loader = tf.data.Dataset.from_tensor_slices((x_train, y_train))
validation_loader = tf.data.Dataset.from_tensor_slices((x_val, y_val))
batch_size = 2
# Augment the on the fly during training.
train_dataset = (
train_loader.shuffle(len(x_train))
.map(train_preprocessing)
.batch(batch_size)
.prefetch(2)
)
# Only rescale.
validation_dataset = (
validation_loader.shuffle(len(x_val))
.map(validation_preprocessing)
.batch(batch_size)
.prefetch(2)
) | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Visualize an augmented CT scan. | import matplotlib.pyplot as plt
data = train_dataset.take(1)
images, labels = list(data)[0]
images = images.numpy()
image = images[0]
print("Dimension of the CT scan is:", image.shape)
plt.imshow(np.squeeze(image[:, :, 30]), cmap="gray")
| examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Since a CT scan has many slices, let's visualize a montage of the slices. |
def plot_slices(num_rows, num_columns, width, height, data):
"""Plot a montage of 20 CT slices"""
data = np.rot90(np.array(data))
data = np.transpose(data)
data = np.reshape(data, (num_rows, num_columns, width, height))
rows_data, columns_data = data.shape[0], data.shape[1]
heights = [slc[0].shape[0] for slc in data]
widths = [slc.shape[1] for slc in data[0]]
fig_width = 12.0
fig_height = fig_width * sum(heights) / sum(widths)
f, axarr = plt.subplots(
rows_data,
columns_data,
figsize=(fig_width, fig_height),
gridspec_kw={"height_ratios": heights},
)
for i in range(rows_data):
for j in range(columns_data):
axarr[i, j].imshow(data[i][j], cmap="gray")
axarr[i, j].axis("off")
plt.subplots_adjust(wspace=0, hspace=0, left=0, right=1, bottom=0, top=1)
plt.show()
# Visualize montage of slices.
# 4 rows and 10 columns for 100 slices of the CT scan.
plot_slices(4, 10, 128, 128, image[:, :, :40]) | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Define a 3D convolutional neural network
To make the model easier to understand, we structure it into blocks.
The architecture of the 3D CNN used in this example
is based on this paper. |
def get_model(width=128, height=128, depth=64):
"""Build a 3D convolutional neural network model."""
inputs = keras.Input((width, height, depth, 1))
x = layers.Conv3D(filters=64, kernel_size=3, activation="relu")(inputs)
x = layers.MaxPool3D(pool_size=2)(x)
x = layers.BatchNormalization()(x)
x = layers.Conv3D(filters=64, kernel_size=3, activation="relu")(x)
x = layers.MaxPool3D(pool_size=2)(x)
x = layers.BatchNormalization()(x)
x = layers.Conv3D(filters=128, kernel_size=3, activation="relu")(x)
x = layers.MaxPool3D(pool_size=2)(x)
x = layers.BatchNormalization()(x)
x = layers.Conv3D(filters=256, kernel_size=3, activation="relu")(x)
x = layers.MaxPool3D(pool_size=2)(x)
x = layers.BatchNormalization()(x)
x = layers.GlobalAveragePooling3D()(x)
x = layers.Dense(units=512, activation="relu")(x)
x = layers.Dropout(0.3)(x)
outputs = layers.Dense(units=1, activation="sigmoid")(x)
# Define the model.
model = keras.Model(inputs, outputs, name="3dcnn")
return model
# Build model.
model = get_model(width=128, height=128, depth=64)
model.summary() | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Train model | # Compile model.
initial_learning_rate = 0.0001
lr_schedule = keras.optimizers.schedules.ExponentialDecay(
initial_learning_rate, decay_steps=100000, decay_rate=0.96, staircase=True
)
model.compile(
loss="binary_crossentropy",
optimizer=keras.optimizers.Adam(learning_rate=lr_schedule),
metrics=["acc"],
)
# Define callbacks.
checkpoint_cb = keras.callbacks.ModelCheckpoint(
"3d_image_classification.h5", save_best_only=True
)
early_stopping_cb = keras.callbacks.EarlyStopping(monitor="val_acc", patience=15)
# Train the model, doing validation at the end of each epoch
epochs = 100
model.fit(
train_dataset,
validation_data=validation_dataset,
epochs=epochs,
shuffle=True,
verbose=2,
callbacks=[checkpoint_cb, early_stopping_cb],
) | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
It is important to note that the number of samples is very small (only 200) and we don't
specify a random seed. As such, you can expect significant variance in the results. The full dataset
which consists of over 1000 CT scans can be found here. Using the full
dataset, an accuracy of 83% was achieved. A variability of 6-7% in the classification
performance is observed in both cases.
Visualizing model performance
Here the model accuracy and loss for the training and the validation sets are plotted.
Since the validation set is class-balanced, accuracy provides an unbiased representation
of the model's performance. | fig, ax = plt.subplots(1, 2, figsize=(20, 3))
ax = ax.ravel()
for i, metric in enumerate(["acc", "loss"]):
ax[i].plot(model.history.history[metric])
ax[i].plot(model.history.history["val_" + metric])
ax[i].set_title("Model {}".format(metric))
ax[i].set_xlabel("epochs")
ax[i].set_ylabel(metric)
ax[i].legend(["train", "val"]) | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
Make predictions on a single CT scan | # Load best weights.
model.load_weights("3d_image_classification.h5")
prediction = model.predict(np.expand_dims(x_val[0], axis=0))[0]
scores = [1 - prediction[0], prediction[0]]
class_names = ["normal", "abnormal"]
for score, name in zip(scores, class_names):
print(
"This model is %.2f percent confident that CT scan is %s"
% ((100 * score), name)
) | examples/vision/ipynb/3D_image_classification.ipynb | keras-team/keras-io | apache-2.0 |
model
load data | %matplotlib inline
import os
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
# define the labels
col_labels=['C2H3', 'C2H6', 'CH2', 'H2CN', 'C2H4', 'H2O2', 'C2H', 'CN',
'heatRelease', 'NCO', 'NNH', 'N2', 'AR', 'psi', 'CO', 'CH4', 'HNCO',
'CH2OH', 'HCCO', 'CH2CO', 'CH', 'mu', 'C2H2', 'C2H5', 'H2', 'T', 'PVs',
'O', 'O2', 'N2O', 'C', 'C3H7', 'CH2(S)', 'NH3', 'HO2', 'NO', 'HCO',
'NO2', 'OH', 'HCNO', 'CH3CHO', 'CH3', 'NH', 'alpha', 'CH3O', 'CO2',
'CH3OH', 'CH2CHO', 'CH2O', 'C3H8', 'HNO', 'NH2', 'HCN', 'H', 'N', 'H2O',
'HCCOH', 'HCNN']
# Taking 0 out
col_labels.remove('AR')
col_labels.remove('heatRelease')
# labels = ['CH4','O2','H2O','CO','CO2','T','PVs','psi','mu','alpha']
labels = ['T','PVs']
# labels = ['T','CH4','O2','CO2','CO','H2O','H2','OH','psi']
# labels = ['CH2OH','HNCO','CH3OH', 'CH2CHO', 'CH2O', 'C3H8', 'HNO', 'NH2', 'HCN']
# labels = np.random.choice(col_labels,20,replace=False).tolist()
# labels.append('PVs')
# labels = col_labels
print(labels)
input_features=['f','pv','zeta']
# read in the data
x_input, y_label, df, in_scaler, out_scaler = read_h5_data('./data/tables_of_fgm.h5',input_features=input_features, labels = labels) | FPV_ANN/notebooks/.ipynb_checkpoints/fgm_nn_inhouse-checkpoint.ipynb | uqyge/combustionML | mit |
model training
gpu training | import keras.backend as K
from keras.callbacks import LearningRateScheduler
import math
def cubic_loss(y_true, y_pred):
return K.mean(K.square(y_true - y_pred)*K.abs(y_true - y_pred), axis=-1)
def coeff_r2(y_true, y_pred):
from keras import backend as K
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 step_decay(epoch):
initial_lrate = 0.001
drop = 0.5
epochs_drop = 200.0
lrate = initial_lrate * math.pow(drop,math.floor((1+epoch)/epochs_drop))
return lrate
lrate = LearningRateScheduler(step_decay)
from keras import optimizers
batch_size = 1024*32
epochs = 60
vsplit = 0.1
loss_type='mse'
adam_op = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999,epsilon=1e-8, decay=0.0, amsgrad=True)
model.compile(loss=loss_type, optimizer=adam_op, metrics=[coeff_r2])
# model.compile(loss=cubic_loss, optimizer=adam_op, metrics=['accuracy'])
# checkpoint (save the best model based validate loss)
!mkdir ./tmp
filepath = "./tmp/weights.best.cntk.hdf5"
checkpoint = ModelCheckpoint(filepath,
monitor='val_loss',
verbose=1,
save_best_only=True,
mode='min',
period=20)
# callbacks_list = [checkpoint]
callbacks_list = [lrate]
# fit the model
history = model.fit(
x_train, y_train,
epochs=epochs,
batch_size=batch_size,
validation_split=vsplit,
verbose=2,
# callbacks=callbacks_list,
shuffle=True)
model.save('trained_fgm_nn.h5') | FPV_ANN/notebooks/.ipynb_checkpoints/fgm_nn_inhouse-checkpoint.ipynb | uqyge/combustionML | mit |
prepare data for plotting
GPU data prepare | from sklearn.metrics import r2_score
# model.load_weights("./tmp/weights.best.cntk.hdf5")
x_test_df = pd.DataFrame(in_scaler.inverse_transform(x_test),columns=input_features)
y_test_df = pd.DataFrame(out_scaler.inverse_transform(y_test),columns=labels)
predict_val = model.predict(x_test,batch_size=1024*8)
predict_df = pd.DataFrame(out_scaler.inverse_transform(predict_val), columns=labels)
test_data=pd.concat([x_test_df,y_test_df],axis=1)
pred_data=pd.concat([x_test_df,predict_df],axis=1)
!rm sim_check.h5
test_data.to_hdf('sim_check.h5',key='test')
pred_data.to_hdf('sim_check.h5',key='pred')
df_test=pd.read_hdf('sim_check.h5',key='test')
df_pred=pd.read_hdf('sim_check.h5',key='pred')
zeta_level=list(set(df_test['zeta']))
zeta_level.sort()
res_sum=pd.DataFrame()
r2s=[]
r2s_i=[]
names=[]
maxs_0=[]
maxs_9=[]
for r2,name in zip(r2_score(df_test,df_pred,multioutput='raw_values'),df_test.columns):
names.append(name)
r2s.append(r2)
maxs_0.append(df_test[df_test['zeta']==zeta_level[0]][name].max())
maxs_9.append(df_test[df_test['zeta']==zeta_level[8]][name].max())
for i in zeta_level:
r2s_i.append(r2_score(df_pred[df_pred['zeta']==i][name],
df_test[df_test['zeta']==i][name]))
res_sum['name']=names
# res_sum['max_0']=maxs_0
# res_sum['max_9']=maxs_9
res_sum['z_scale']=[m_9/(m_0+1e-20) for m_9,m_0 in zip(maxs_9,maxs_0)]
# res_sum['r2']=r2s
tmp=np.asarray(r2s_i).reshape(-1,10)
for idx,z in enumerate(zeta_level):
res_sum['r2s_'+str(z)]=tmp[:,idx]
res_sum[3:]
no_drop=res_sum[3:]
no_drop | FPV_ANN/notebooks/.ipynb_checkpoints/fgm_nn_inhouse-checkpoint.ipynb | uqyge/combustionML | mit |
The optimization problem
The problem we are considering is a mathematical one
<img src="cone.png" width=500px/>
Decisions: r in [0, 10] cm; h in [0, 20] cm
Objectives: minimize S, T
Constraints: V > 200cm<sup>3</sup> | # Few Utility functions
def say(*lst):
"""
Print whithout going to new line
"""
print(*lst, end="")
sys.stdout.flush()
def random_value(low, high, decimals=2):
"""
Generate a random number between low and high.
decimals incidicate number of decimal places
"""
return round(random.uniform(low, high),decimals)
def gt(a, b): return a > b
def lt(a, b): return a < b
def shuffle(lst):
"""
Shuffle a list
"""
random.shuffle(lst)
return lst
class Decision(O):
"""
Class indicating Decision of a problem
"""
def __init__(self, name, low, high):
"""
@param name: Name of the decision
@param low: minimum value
@param high: maximum value
"""
O.__init__(self, name=name, low=low, high=high)
class Objective(O):
"""
Class indicating Objective of a problem
"""
def __init__(self, name, do_minimize=True):
"""
@param name: Name of the objective
@param do_minimize: Flag indicating if objective has to be minimized or maximized
"""
O.__init__(self, name=name, do_minimize=do_minimize)
class Point(O):
"""
Represents a member of the population
"""
def __init__(self, decisions):
O.__init__(self)
self.decisions = decisions
self.objectives = None
def __hash__(self):
return hash(tuple(self.decisions))
def __eq__(self, other):
return self.decisions == other.decisions
def clone(self):
new = Point(self.decisions)
new.objectives = self.objectives
return new
class Problem(O):
"""
Class representing the cone problem.
"""
def __init__(self):
O.__init__(self)
# TODO 2: Code up decisions and objectives below for the problem
# using the auxilary classes provided above.
self.decisions = None
self.objectives = None
radius = Decision('radius', 0, 10)
height = Decision('height', 0, 20)
self.decisions = [radius, height]
s = Objective('surface')
t = Objective('total area')
self.objectives = [s,t]
@staticmethod
def evaluate(point):
[r, h] = point.decisions
point.objectives = None
# TODO 3: Evaluate the objectives S and T for the point.
l = (r**2 + h**2)**0.5
S = pi * r * l
T = S + pi * r**2
point.objectives = [S, T]
return point.objectives
@staticmethod
def is_valid(point):
[r, h] = point.decisions
# TODO 4: Check if the point has valid decisions
V = pi*(r**2)*h/3
return V > 200
def generate_one(self):
# TODO 5: Generate a valid instance of Point.
while True:
point = Point([random_value(d.low, d.high) for d in self.decisions])
if Problem.is_valid(point):
return point
cone = Problem()
point = cone.generate_one()
cone.evaluate(point)
print(point) | code/5/WS1/tchhabr.ipynb | tarunchhabra26/fss16dst | apache-2.0 |
Great. Now that the class and its basic methods is defined, we move on to code up the GA.
Population
First up is to create an initial population. | def populate(problem, size):
population = []
# TODO 6: Create a list of points of length 'size'
return [problem.generate_one() for _ in xrange(size)]
print (populate(cone,5))
| code/5/WS1/tchhabr.ipynb | tarunchhabra26/fss16dst | apache-2.0 |
Crossover
We perform a single point crossover between two points | def crossover(mom, dad):
# TODO 7: Create a new point which contains decisions from
# the first half of mom and second half of dad
n = len(mom.decisions)
return Point(mom.decisions[:n//2] + dad.decisions[n//2:])
pop = populate(cone,5)
crossover(pop[0], pop[1]) | code/5/WS1/tchhabr.ipynb | tarunchhabra26/fss16dst | apache-2.0 |
Mutation
Randomly change a decision such that | def mutate(problem, point, mutation_rate=0.01):
# TODO 8: Iterate through all the decisions in the problem
# and if the probability is less than mutation rate
# change the decision(randomly set it between its max and min).
for i, d in enumerate(problem.decisions):
if random.random() < mutation_rate:
point.decisions[i] = random_value(d.low, d.high)
return point
print (mutate(cone,point,0.1))
obs = populate(cone,5)
print (obs) | code/5/WS1/tchhabr.ipynb | tarunchhabra26/fss16dst | apache-2.0 |
Fitness Evaluation
To evaluate fitness between points we use binary domination. Binary Domination is defined as follows:
* Consider two points one and two.
* For every decision o and t in one and two, o <= t
* Atleast one decision o and t in one and two, o == t
Note: Binary Domination is not the best method to evaluate fitness but due to its simplicity we choose to use it for this workshop. | def bdom(problem, one, two):
"""
Return if one dominates two
"""
objs_one = problem.evaluate(one)
objs_two = problem.evaluate(two)
if (one == two):
return False
dominates = False
# TODO 9: Return True/False based on the definition
# of bdom above.
first = True
second = False
for i,_ in enumerate(problem.objectives):
if ((first is True) & gt(one.objectives[i], two.objectives[i])):
first = False
elif (not second & (one.objectives[i] is not two.objectives[i])):
second = True
dominates = first & second
return dominates
print (bdom(cone,obs[4],obs[4])) | code/5/WS1/tchhabr.ipynb | tarunchhabra26/fss16dst | apache-2.0 |
Fitness and Elitism
In this workshop we will count the number of points of the population P dominated by a point A as the fitness of point A. This is a very naive measure of fitness since we are using binary domination.
Few prominent alternate methods are
1. Continuous Domination - Section 3.1
2. Non-dominated Sort
3. Non-dominated Sort + Niching
Elitism: Sort points with respect to the fitness and select the top points. | def fitness(problem, population, point):
dominates = 0
# TODO 10: Evaluate fitness of a point.
# For this workshop define fitness of a point
# as the number of points dominated by it.
# For example point dominates 5 members of population,
# then fitness of point is 5.
for pop in population:
if bdom(problem, point, pop):
dominates += 1
return dominates
def elitism(problem, population, retain_size):
# TODO 11: Sort the population with respect to the fitness
# of the points and return the top 'retain_size' points of the population
fit_pop = [fitness(cone,population,pop) for pop in population]
population = [pop for _,pop in sorted(zip(fit_pop,population), reverse = True)]
return population[:retain_size] | code/5/WS1/tchhabr.ipynb | tarunchhabra26/fss16dst | apache-2.0 |
Load the pretrained weights into the network : | params = pickle.load(open('./data/googlenet/blvc_googlenet.pkl', 'rb'), encoding='iso-8859-1')
model_param_values = params['param values']
#classes = params['synset words']
lasagne.layers.set_all_param_values(net_output_layer, model_param_values)
IMAGE_W=224
print("Loaded Model parameters") | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Executing the cell below will iterate through the images in the ./images/art-style/photos directory, so you can choose the one you want | photo_i += 1
photo = plt.imread(photos[photo_i % len(photos)])
photo_rawim, photo = googlenet.prep_image(photo)
plt.imshow(photo_rawim) | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Executing the cell below will iterate through the images in the ./images/art-style/styles directory, so you can choose the one you want | style_i += 1
art = plt.imread(styles[style_i % len(styles)])
art_rawim, art = googlenet.prep_image(art)
plt.imshow(art_rawim) | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
This defines various measures of difference that we'll use to compare the current output image with the original sources. | def plot_layout(combined):
def no_axes():
plt.gca().xaxis.set_visible(False)
plt.gca().yaxis.set_visible(False)
plt.figure(figsize=(9,6))
plt.subplot2grid( (2,3), (0,0) )
no_axes()
plt.imshow(photo_rawim)
plt.subplot2grid( (2,3), (1,0) )
no_axes()
plt.imshow(art_rawim)
plt.subplot2grid( (2,3), (0,1), colspan=2, rowspan=2 )
no_axes()
plt.imshow(combined, interpolation='nearest')
plt.tight_layout()
def gram_matrix(x):
x = x.flatten(ndim=3)
g = T.tensordot(x, x, axes=([2], [2]))
return g
def content_loss(P, X, layer):
p = P[layer]
x = X[layer]
loss = 1./2 * ((x - p)**2).sum()
return loss
def style_loss(A, X, layer):
a = A[layer]
x = X[layer]
A = gram_matrix(a)
G = gram_matrix(x)
N = a.shape[1]
M = a.shape[2] * a.shape[3]
loss = 1./(4 * N**2 * M**2) * ((G - A)**2).sum()
return loss
def total_variation_loss(x):
return (((x[:,:,:-1,:-1] - x[:,:,1:,:-1])**2 + (x[:,:,:-1,:-1] - x[:,:,:-1,1:])**2)**1.25).sum() | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Here are the GoogLeNet layers that we're going to pay attention to : | layers = [
# used for 'content' in photo - a mid-tier convolutional layer
'inception_4b/output',
# used for 'style' - conv layers throughout model (not same as content one)
'conv1/7x7_s2', 'conv2/3x3', 'inception_3b/output', 'inception_4d/output',
]
#layers = [
# # used for 'content' in photo - a mid-tier convolutional layer
# 'pool4/3x3_s2',
#
# # used for 'style' - conv layers throughout model (not same as content one)
# 'conv1/7x7_s2', 'conv2/3x3', 'pool3/3x3_s2', 'inception_5b/output',
#]
layers = {k: net[k] for k in layers} | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Precompute layer activations for photo and artwork
This takes ~ 20 seconds | input_im_theano = T.tensor4()
outputs = lasagne.layers.get_output(layers.values(), input_im_theano)
photo_features = {k: theano.shared(output.eval({input_im_theano: photo}))
for k, output in zip(layers.keys(), outputs)}
art_features = {k: theano.shared(output.eval({input_im_theano: art}))
for k, output in zip(layers.keys(), outputs)}
# Get expressions for layer activations for generated image
generated_image = theano.shared(floatX(np.random.uniform(-128, 128, (1, 3, IMAGE_W, IMAGE_W))))
gen_features = lasagne.layers.get_output(layers.values(), generated_image)
gen_features = {k: v for k, v in zip(layers.keys(), gen_features)} | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Define the overall loss / badness function | losses = []
# content loss
cl = 10 /1000.
losses.append(cl * content_loss(photo_features, gen_features, 'inception_4b/output'))
# style loss
sl = 20 *1000.
losses.append(sl * style_loss(art_features, gen_features, 'conv1/7x7_s2'))
losses.append(sl * style_loss(art_features, gen_features, 'conv2/3x3'))
losses.append(sl * style_loss(art_features, gen_features, 'inception_3b/output'))
losses.append(sl * style_loss(art_features, gen_features, 'inception_4d/output'))
#losses.append(sl * style_loss(art_features, gen_features, 'inception_5b/output'))
# total variation penalty
vp = 0.01 /1000. /1000.
losses.append(vp * total_variation_loss(generated_image))
total_loss = sum(losses) | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
The Famous Symbolic Gradient operation | grad = T.grad(total_loss, generated_image) | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Get Ready for Optimisation by SciPy | # Theano functions to evaluate loss and gradient - takes around 1 minute (!)
f_loss = theano.function([], total_loss)
f_grad = theano.function([], grad)
# Helper functions to interface with scipy.optimize
def eval_loss(x0):
x0 = floatX(x0.reshape((1, 3, IMAGE_W, IMAGE_W)))
generated_image.set_value(x0)
return f_loss().astype('float64')
def eval_grad(x0):
x0 = floatX(x0.reshape((1, 3, IMAGE_W, IMAGE_W)))
generated_image.set_value(x0)
return np.array(f_grad()).flatten().astype('float64') | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Initialize with the original photo, since going from noise (the code that's commented out) takes many more iterations. | generated_image.set_value(photo)
#generated_image.set_value(floatX(np.random.uniform(-128, 128, (1, 3, IMAGE_W, IMAGE_W))))
x0 = generated_image.get_value().astype('float64')
iteration=0 | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Optimize all those losses, and show the image
To refine the result, just keep hitting 'run' on this cell (each iteration is about 60 seconds) : | t0 = time.time()
scipy.optimize.fmin_l_bfgs_b(eval_loss, x0.flatten(), fprime=eval_grad, maxfun=40)
x0 = generated_image.get_value().astype('float64')
iteration += 1
if False:
plt.figure(figsize=(8,8))
plt.imshow(googlenet.deprocess(x0), interpolation='nearest')
plt.axis('off')
plt.text(270, 25, '# {} in {:.1f}sec'.format(iteration, (float(time.time() - t0))), fontsize=14)
else:
plot_layout(googlenet.deprocess(x0))
print('Iteration {}, ran in {:.1f}sec'.format(iteration, float(time.time() - t0))) | notebooks/2-CNN/6-StyleTransfer/2-Art-Style-Transfer-googlenet_theano.ipynb | mdda/fossasia-2016_deep-learning | mit |
Note about slicing columns from a Numpy matrix
If you want to extract a column i from a Numpy matrix A and keep it as a column vector, you need to use the slicing notation, A[:, i:i+1]. Not doing so can lead to subtle bugs. To see why, compare the following slices. | A = np.array ([[1, 2, 3],
[4, 5, 6],
[7, 8, 9]
], dtype=float)
print "A[:, :] ==\n", A
print "\nA[:, 0] ==\n", A[:, 0]
print "\nA[:, 2:3] == \n", A[:, 2:3]
print "\nAdd columns 0 and 2?"
a0 = A[:, 0]
a1 = A[:, 2:3]
print a0 + a1 | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Sample data: Rock lobsters!
As a concrete example of a classification task, consider the results of this experiment. Some marine biologists took a bunch of lobsters of varying sizes (size being a proxy for stage of development), and then tethered and exposed these lobsters to a variety of predators. The outcome that they measured is whether the lobsters survived or not.
In this case, the data consists of a set of points, one point per lobster, where there is a single predictor (size) and the response is whether the lobsters survived (label "1") or died (label "0").
For the original paper, see this link. I can only imagine that this image is what marine biologists look like when experimenting with lobsters.
Here is a plot of the raw data. | # http://www.stat.ufl.edu/~winner/data/lobster_survive.txt
df_lobsters = pd.read_table ('http://www.stat.ufl.edu/~winner/data/lobster_survive.dat',
sep=r'\s+', names=['CarapaceLen', 'Survived'])
display (df_lobsters.head ())
print "..."
display (df_lobsters.tail ())
sns.violinplot (x="Survived", y="CarapaceLen",
data=df_lobsters, inner="quart") | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Although the classes are distinct in the aggregate, where the median carapace (outer shell) length is around 36 mm for the lobsters that died and 42 mm for those that survived, they are not cleanly separable.
Notation
To develop some intuition and a method, let's now turn to a more general setting and work on synthetic data sets.
Let the data consist of $m$ data points, where each point is $d$-dimensional. Each dimension corresponds to some continuously-valued predictor. In addition, each data point will have a binary label, whose value is either 0 or 1.
Denote each point by an augumented vector, $x_i$, such that
$$
\begin{array}{rcl}
x_i
& \equiv &
\left(\begin{array}{c}
1 \
x_{i,1} \
x_{i,2} \
\vdots \
x_{i,d}
\end{array}\right)
.
\end{array}
$$
That is, the point is the $d$ coordinates augmented by an initial dummy coordinate whose value is 1. This convention is similar to what we did in linear regression.
We can also stack these points as rows of a matrix, $X$, again, just as we did in regression:
$$
\begin{array}{rcl}
X \equiv
\left(\begin{array}{c}
x_0^T \
x_1^T \
\vdots \
x_{m-1}^T
\end{array}\right)
& = &
\left(\begin{array}{ccccc}
1 & x_{0,1} & x_{0,2} & \cdots & x_{0,d} \
1 & x_{1,1} & x_{1,2} & \cdots & x_{1,d} \
& & & \vdots & \
1 & x_{m-1,1} & x_{m-1,2} & \cdots & x_{m-1,d} \
\end{array}\right).
\end{array}
$$
We will take the labels to be a binary column vector, $l \equiv \left(l_0, l_1, \ldots, l_{m-1}\right)^T$.
An example
We've pre-generated a synethetic data set consisting of labeled data points. Let's download and inspect it, first as a table and then visually. | df = pd.read_csv ('http://vuduc.org/cse6040/logreg_points_train.csv')
display (df.head ())
print "..."
display (df.tail ()) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Next, let's extract the coordinates as a Numpy matrix of points and the labels as a Numpy column vector labels. Mathematically, the points matrix corresponds to $X$ and the labels vector corresponds to $l$. | points = np.insert (df.as_matrix (['x_1', 'x_2']), 0, 1.0, axis=1)
labels = df.as_matrix (['label'])
print "First and last 5 points:\n", '='*23, '\n', points[:5], '\n...\n', points[-5:], '\n'
print "First and last 5 labels:\n", '='*23, '\n', labels[:5], '\n...\n', labels[-5:], '\n' | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Next, let's plot the data as a scatter plot using Plotly. To do so, we need to create separate traces, one for each cluster. Below, we've provided you with a function, make_2d_scatter_traces(), which does exactly that, given a labeled data set as a (points, labels) pair. | def assert_points_2d (points):
"""Checks the dimensions of a given point set."""
assert type (points) is np.ndarray
assert points.ndim == 2
assert points.shape[1] == 3
def assert_labels (labels):
"""Checks the type of a given set of labels (must be integral)."""
assert labels is not None
assert (type (labels) is np.ndarray) or (type (labels) is list)
def extract_clusters (points, labels):
"""
Given a list or array of labeled augmented points, this
routine returns a pair of lists, (C[0:k], L[0:k]), where
C[i] is an array of all points whose labels are L[i].
"""
assert_points_2d (points)
assert_labels (labels)
id_label_pairs = list (enumerate (set (labels.flatten ())))
labels_map = dict ([(v, i) for (i, v) in id_label_pairs])
# Count how many points belong to each cluster
counts = [0] * len (labels_map)
for l in labels.flatten ():
counts[labels_map[l]] += 1
# Allocate space for each cluster
clusters = [np.zeros ((k, 3)) for k in counts]
# Separate the points by cluster
counts = [0] * len (labels_map)
for (x, l) in zip (points, labels.flatten ()):
l_id = labels_map[l]
k = counts[l_id]
clusters[l_id][k, :] = x
counts[l_id] += 1
# Generate cluster labels
cluster_labels = [None] * len (labels_map)
for (l, i) in labels_map.items ():
cluster_labels[i] = l
return (clusters, cluster_labels)
def make_2d_scatter_traces (points, labels=None):
"""
Given an augmented point set, possibly labeled,
returns a list Plotly-compatible marker traces.
"""
assert_points_2d (points)
traces = []
if labels is None:
traces.append (Scatter (x=points[:, 1:2], y=points[:, 2:3], mode='markers'))
else:
assert_labels (labels)
(clusters, cluster_labels) = extract_clusters (points, labels)
for (c, l) in zip (clusters, cluster_labels):
traces.append (Scatter (x=c[:, 1:2], y=c[:, 2:3],
mode='markers',
name="%s" % str (l)))
return traces
print "Number of points:", len (points)
traces = make_2d_scatter_traces (points, labels)
py.iplot (traces) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Linear discriminants
Suppose you think that the boundary between the two clusters may be represented by a line. For the synthetic data example above, I hope you'll agree that such a model is not a terrible one.
This line is referred to as a linear discriminant. Any point $x$ on this line may be described by $\theta^T x$, where $\theta$ is a vector of coefficients:
$$
\begin{array}{rcl}
\theta
& \equiv &
\left(\begin{array}{c} \theta_0 \ \theta_1 \ \vdots \ \theta_d \end{array}\right)
.
\
\end{array}
$$
For example, consider the case of 2-D points ($d=2$): the condition that $\theta^T x = 0$ means that
$$
\begin{array}{rrcl}
&
\theta^T x = 0
& = & \theta_0 + \theta_1 x_1 + \theta_2 x_2 \
\implies
& x_2
& = & -\frac{\theta_0}{\theta_2} - \frac{\theta_1}{\theta_2} x_1.
\end{array}
$$
So that describes points on the line. However, given any point $x$ in the $d$-dimensional space that is not on the line, $\theta^T x$ still produces a value: that value will be positive on one side of the line ($\theta^T x > 0$) or negative on the other ($\theta^T x < 0$).
Consequently, here is one simple way to use the linear discriminant function $\theta^T x$ to generate a label: just reinterpret its sign! In more mathematical terms, the function that converts, say, a positive value to the label "1" and all other values to the label "0" is called the heaviside function:
$$
\begin{array}{rcl}
H(y) & \equiv & \left{\begin{array}{ll}
1 & \mathrm{if}\ y > 0
\
0 & \mathrm{if}\ y \leq 0
\end{array}\right..
\end{array}
$$
Exercise. This exercise has three parts.
1) Given the a $m \times (d+1)$ matrix of augmented points (i.e., the $X$ matrix) and the vector $\theta$, implement a function to compute the value of the linear discriminant at each point. That is, the function should return a (column) vector $y$ where the $y_i = \theta^T x_i$.
2) Implement the heaviside function, $H(y)$. Your function should allow for an arbitrary matrix of input values, and should apply the heaviside function elementwise.
Hint: Consider what Numpy's sign() function produces, and transform the result accordingly.
3) For the synthetic data you loaded above, determine a value of $\theta$ for which $H(\theta^T x)$ "best" separates the two clusters. To help you out, we've provided some Plotly code that draws the discriminant boundary and also applies $H(\theta^T x)$ to each point, coloring the point by whether it is correctly classified. (The code also prints the number of correcty classified points.) So, you just need to try different values of $\theta$ until you find something that is "close."
Hint: We found a line that commits just 5 errors, out of 375 possible points. | def lin_discr (X, theta):
# @YOUSE: Part 1 -- Complete this function.
pass
def heaviside (Y):
# @YOUSE: Part 2 -- Complete this function
pass | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
The following is the code to generate the plot; look for the place to try different values of $\theta$ a couple of code cells below. | def heaviside_int (Y):
"""Evaluates the heaviside function, but returns integer values."""
return heaviside (Y).astype (dtype=int)
def assert_discriminant (theta, d=2):
"""
Verifies that the given coefficients correspond to a
d-dimensional linear discriminant ($\theta$).
"""
assert len (theta) == (d+1)
def gen_lin_discr_labels (points, theta, fun=heaviside_int):
"""
Given a set of points and the coefficients of a linear
discriminant, this function returns a set of labels for
the points with respect to this discriminant.
"""
assert_points_2d (points)
assert_discriminant (theta)
score = lin_discr (points, theta)
labels = fun (score)
return labels
def gen_lin_discr_trace (points, theta, name='Discriminant'):
"""
Given a set of points and the coefficients of a linear
discriminant, this function returns a set of Plotly
traces that show how the points are classified as well
as the location of the discriminant boundary.
"""
assert_points_2d (points)
assert_discriminant (theta)
x1 = [min (points[:, 1]), max (points[:, 1])]
m = -theta[1] / theta[2]
b = -theta[0] / theta[2]
x2 = [(b + m*x) for x in x1]
return Scatter (x=x1, y=x2, mode='lines', name=name)
def np_row_vec (init_list):
"""Generates a Numpy-compatible row vector."""
return np.array (init_list, order='F', ndmin=2)
def np_col_vec (init_list):
"""Generates a Numpy-compatible column vector."""
return np_row_vec (init_list).T
def gen_labels_part3 (points, labels, theta):
your_labels = gen_lin_discr_labels (points, theta)
return (labels == your_labels)
# @YOUSE: Part 3 -- Select parameters for theta!
theta = np_col_vec ([0., -1., 3.])
# Generate 0/1 labels for your discriminant:
is_correct = gen_labels_part3 (points, labels, theta)
print "Number of misclassified points:", (len (points) - sum (is_correct))[0]
print "\n(Run the code cell below to visualize the results.)"
# Visually inspect the above results
traces = make_2d_scatter_traces (points, is_correct)
traces.append (gen_lin_discr_trace (points, theta))
# Plot it!
layout = Layout (xaxis=dict (range=[-1.25, 2.25]),
yaxis=dict (range=[-3.25, 2.25]))
fig = Figure (data=traces, layout=layout)
py.iplot (fig) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
An alternative linear discriminant: the logistic or "sigmoid" function
The heaviside function, $H(\theta^T x)$, enforces a sharp boundary between classes around the $\theta^T x=0$ line. The following code produces a contour plot to show this effect. | # Use Numpy's handy meshgrid() to create a regularly-spaced grid of values.
# http://docs.scipy.org/doc/numpy/reference/generated/numpy.meshgrid.html
x1 = np.linspace (-2., +2., 100)
x2 = np.linspace (-2., +2., 100)
x1_grid, x2_grid = np.meshgrid (x1, x2)
h_grid = heaviside (theta[0] + theta[1]*x1_grid + theta[2]*x2_grid)
trace_grid = Contour (x=x1, y=x2, z=h_grid)
py.iplot ([trace_grid]) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
However, as the lobsters example suggests, real data are not likely to be cleanly separable, especially when the number of features we have at our disposal is relatively small.
Since the labels are binary, a natural idea is to give the classification problem a probabilistic interpretation. The logistic function provides at least one way to do so:
$$
\begin{array}{rcl}
G(y) & \equiv & \frac{1}{1 + e^{-y}}
\end{array}
$$
This function is also sometimes called the logit or sigmoid function.
The logistic function takes any value in the range $(-\infty, +\infty)$ and produces a value in the range $(0, 1)$. Thus, given a value $x$, we can interpret it as a conditional probability that the label is 1.
Exercise. Consider a set of 1-D points generated by a mixture of Gaussians. That is, suppose that there are two Gaussian distributions over the 1-dimensional variable, $x \in (-\infty, +\infty)$, that have the same variance ($\sigma^2$) but different means ($\mu_0$ and $\mu_1$). Show that the conditional probability of observing a point labeled "1" given $x$ may be written as,
$$\mathrm{Pr}\left[l=1\,|\,x\right]
\propto \displaystyle \frac{1}{1 + e^{-(\theta_0 + \theta_1 x)}},$$
for a suitable definition of $\theta_0$ and $\theta_1$. To carry out this computation, recall Bayes's rule (also: Bayes's theorem):
$$
\begin{array}{rcl}
\mathrm{Pr}[l=1\,|\,x]
& = &
\dfrac{\mathrm{Pr}[x\,|\,l=1] \, \mathrm{Pr}[l=1]}
{\mathrm{Pr}[x\,|\,l=0] \, \mathrm{Pr}[l=0]
+ \mathrm{Pr}[x\,|\,l=1] \, \mathrm{Pr}[l=1]
}.
\end{array}
$$
You may assume the prior probabilities of observing a 0 or 1 are given by $\mathrm{Pr}[l=0] \equiv p_0$ and $\mathrm{Pr}[l=1] \equiv p_1$.
Time and interest permitting, we'll solve this exercise on the whiteboard.
Exercise. Implement the logistic function. Inspect the resulting plot of $G(y)$ in 1-D and then the contour plot of $G(\theta^T{x})$. Your function should accept a Numpy matrix of values, Y, and apply the sigmoid elementwise. | def logistic (Y):
# @YOUSE: Implement the logistic function G(y) here
pass | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Plot of your implementation in 1D: | x_logit_1d = np.linspace (-6.0, +6.0, 101)
y_logit_1d = logistic (x_logit_1d)
trace_logit_1d = Scatter (x=x_logit_1d, y=y_logit_1d)
py.iplot ([trace_logit_1d]) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Contour plot of your function: | g_grid = logistic (theta[0] + theta[1]*x1_grid + theta[2]*x2_grid)
trace_logit_grid = Contour (x=x1, y=x2, z=g_grid)
py.iplot ([trace_logit_grid]) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Exercise. Verify the following properties of the logistic function, $G(y)$.
$$
\begin{array}{rcll}
G(y)
& = & \frac{e^y}{e^y + 1}
& \mathrm{(P1)} \
G(-y)
& = & 1 - G(y)
& \mathrm{(P2)} \
\dfrac{dG}{dy}
& = & G(y) G(-y)
& \mathrm{(P3)} \
{\dfrac{d}{dy}} {\left[ \ln G(y) \right]}
& = & G(-y)
& \mathrm{(P4)} \
{\dfrac{d}{dy}} {\ln \left[ 1 - G(y) \right]}
& = & -G(y)
& \mathrm{(P5)}
\end{array}
$$
Determining $\theta$ via Maximum Likelihood Estimation
Previously, you determined $\theta$ for our synthetic dataset experimentally. Can you compute a good $\theta$ automatically? One of the standard techniques in statistics is to perform a maximum likelihood estimation (MLE) of a model's parameters, $\theta$.
Indeed, MLE is basis for the "statistical" way to derive the normal equations in the case of linear regression, though that is of course not how we encountered it in this class.
"Likelihood" as an objective function
MLE derives from the following idea. Consider the joint probability of observing all of the labels, given the points and the parameters, $\theta$:
$$
\mathrm{Pr}[l\,|\,X, \theta].
$$
Suppose these observations are independent and identically distributed (i.i.d.). Then the joint probability can be factored as the product of individual probabilities,
$$
\begin{array}{rcl}
\mathrm{Pr}[l\,|\,X,\theta] = \mathrm{Pr}[l_0, \ldots, l_{m-1}\,|\,x_0, \ldots, x_{m-1}, \theta]
& = & \mathrm{Pr}[l_0\,|\,x_0, \theta] \cdots \mathrm{Pr}[l_{m-1}\,|\,x_{m-1}, \theta] \
& = & \displaystyle \prod_{i=0}^{m-1} \mathrm{Pr}[l_i\,|\,x_i,\theta].
\end{array}
$$
The maximum likelihood principle says that you should try to choose a parameter $\theta$ that maximizes the chances ("likelihood") of seeing these particular observations. Thus, we can simply reinterpret the preceding probability as an objective function to optimize. Mathematically, it is equivalent and convenient to consider the logarithm of the likelihood, or log-likelihood, as the objective function, defining it by,
$$
\begin{array}{rcl}
\mathcal{L}(\theta; l, X)
& \equiv &
\log \left{ \displaystyle \prod_{i=0}^{m-1} \mathrm{Pr}[l_i\,|\,x_i,\theta] \right} \
& = &
\displaystyle \sum_{i=0}^{m-1} \log \mathrm{Pr}[l_i\,|\,x_i,\theta].
\end{array}
$$
We are using the symbol $\log$, which could be taken in any convenient base, such as the natural logarithm ($\ln y$) or the information theoretic base-two logarithm ($\log_2 y$).
The MLE procedure then consists of two steps:
For the problem at hand, determine a suitable choice for $\mathrm{Pr}[l_i\,|\,x_i,\theta]$.
Run any optimization procedure to find the $\theta$ that maximizes $\mathcal{L}(\theta; l, X)$.
Example: Logistic regression
Let's say you have decided that the logistic function, $G(\theta^T x_i)$, is a good model of the probability of producing a label $l_i$ given the point $x_i$. Under the i.i.d. assumption, we can interpret the label $l_i$ as being the result of a Bernoulli trial (e.g., a biased coin flip), where the probability of success ($l_i=1$) is defined as $g_i = g_i(\theta) \equiv G(\theta^T x_i)$. Thus,
$$
\begin{array}{rcl}
\mathrm{Pr}[l_i \, | \, x_i, \theta]
& \equiv & g_i^{l_i} \cdot \left(1 - g_i\right)^{1 - l_i}.
\end{array}
$$
The log-likelihood in turn becomes,
$$
\begin{array}{rcl}
\mathcal{L}(\theta; l, X)
& = & \displaystyle
\sum_{i=0}^{m-1} l_i \log g_i + (1-l_i) \log (1-g_i) \
& = &
l^T \log g + (1-l)^T \log (1-g),
\end{array}
$$
where $g \equiv (g_0, g_1, \ldots, g_{m-1})^T$.
Optimizing the log-likelihood via gradient (steepest) ascent
To optimize the log-likelihood with respect to the parameters, $\theta$, you'd like to do the moral equivalent of taking its derivative, setting it to zero, and then solving for $\theta$.
For example, recall that in the case of linear regression via least squares minimization, carrying out this process produced an analytic solution for the parameters, which was to solve the normal equations.
Unfortunately, for logistic regression---or for most log-likelihoods you are likely to ever write down---you cannot usually derive an analytic solution. Therefore, you will need to resort to numerical optimization procedures.
The simplest such procedure is gradient ascent (or steepest ascent), in the case of maximizing some function; if instead you are minimizing the function, then the equivalent procedure is gradient (steepest) descent. The idea is to start with some guess, compute the derivative of the objective function at that guess, and then move in the direction of steepest descent. As it happens, the direction of steepest descent is given by the gradient. More formally, the procedure applied to the log-likelihood is:
Start with some initial guess, $\theta(0)$.
At each iteration $t \geq 0$ of the procedure, let $\theta(t)$ be the current guess.
Compute the direction of steepest descent by evaluating the gradient, $\Delta_t \equiv \nabla_{\theta(t)} \left{\mathcal{L}(\theta(t); l, X)\right}$.
Take a step in the direction of the gradient, $\theta(t+1) \leftarrow \theta(t) + \phi \Delta_t$, where $\phi$ is a suitably chosen fudge factor.
This procedure should smell eerily like the one in Lab 24! And just as in Lab 24, the tricky bit is how to choose $\phi$, the principled choice of which we will defer until another lab.
One additional and slight distinction between this procedure and the Lab 24 procedure is that here we are optimizing using the full dataset, rather than processing data points one at a time. (That is, the step iteration variable $t$ used above is not used in exactly the same way as the step iteration $k$ was used in Lab 24.)
Another question is, how do we know this procedure will converge to the global maximum, rather than, say, a local maximum? For that you need a deeper analysis of a specific $\mathcal{L}(\theta; l, X)$, to show, for instance, that it is convex in $\theta$.
Example: A gradient ascent algorithm for logistic regression
Let's apply the gradient ascent procedure to the logistic regression problem, in order to determine a good $\theta$.
Exercise. Show the following:
$$
\begin{array}{rcl}
\nabla_\theta \left{\mathcal{L}(\theta; l, X)\right}
& = & X^T \left[ l - G(X \cdot \theta)\right].
\end{array}
$$
Exercise. Implement the gradient ascent procedure to determine $\theta$, and try it out on the sample data.
In your solution, we'd like you to store all guesses in the matrix thetas, so that you can later see how the $\theta(t)$ values evolve. To extract a particular column t, use the notation, theta[:, t:t+1]. This notation is necessary to preserve the "shape" of the column as a column vector. | MAX_STEP = 100
PHI = 0.1
# Get the data coordinate matrix, X, and labels vector, l
X = points
l = labels.astype (dtype=float)
# Store *all* guesses, for subsequent analysis
thetas = np.zeros ((3, MAX_STEP+1))
for t in range (MAX_STEP):
# @YOUSE: Fill in this code
pass
print "Your (hand) solution:", theta.T.flatten ()
print "Computed solution:", thetas[:, MAX_STEP]
theta_mle = thetas[:, MAX_STEP:]
# Generate 0/1 labels for computed discriminant:
is_correct_mle = gen_labels_part3 (points, labels, theta_mle)
print "Number of misclassified points using MLE:", (len (points) - sum (is_correct_mle))[0]
print "\n(Run the code cell below to visualize the results.)"
# Visually inspect the above results
traces_mle = make_2d_scatter_traces (points, is_correct_mle)
traces_mle.append (gen_lin_discr_trace (points, theta_mle))
# Plot it!
layout_mle = Layout (xaxis=dict (range=[-1.25, 2.25]),
yaxis=dict (range=[-3.25, 2.25]))
fig_mle = Figure (data=traces_mle, layout=layout_mle)
py.iplot (fig_mle) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
Exercise. Make a contour plot of the log-likelihood and draw the trajectory taken by the $\theta(t)$ values laid on top of it. | def log_likelihood (theta, l, X):
# @YOUSE: Complete this function to evaluate the log-likelihood
pass
n1_ll = 100
x1_ll = np.linspace (-20., 0., n1_ll)
n2_ll = 100
x2_ll = np.linspace (-20., 0., n2_ll)
x1_ll_grid, x2_ll_grid = np.meshgrid (x1_ll, x2_ll)
ll_grid = np.zeros ((n1_ll, n2_ll))
# @YOUSE: Write some code to compute ll_grid, which the following code cell visualizes
trace_ll_grid = Contour (x=x1_ll, y=x2_ll, z=ll_grid)
trace_thetas = Scatter (x=thetas[1, :], y=thetas[2, :], mode='markers+lines')
py.iplot ([trace_ll_grid, trace_thetas]) | 25--logreg.ipynb | rvuduc/cse6040-ipynbs | bsd-3-clause |
<div class="alert alert-success">
<b>EXERCISE</b>: Using groupby(), plot the number of films that have been released each decade in the history of cinema.
</div>
<div class="alert alert-success">
<b>EXERCISE</b>: Use groupby() to plot the number of "Hamlet" films made each decade.
</div>
<div class="alert alert-success">
<b>EXERCISE</b>: How many leading (n=1) roles were available to actors, and how many to actresses, in each year of the 1950s?
</div>
<div class="alert alert-success">
<b>EXERCISE</b>: Use groupby() to determine how many roles are listed for each of The Pink Panther movies.
</div>
<div class="alert alert-success">
<b>EXERCISE</b>: List, in order by year, each of the films in which Frank Oz has played more than 1 role.
</div>
<div class="alert alert-success">
<b>EXERCISE</b>: List each of the characters that Frank Oz has portrayed at least twice.
</div>
Transforms
Sometimes you don't want to aggregate the groups, but transform the values in each group. This can be achieved with transform:£ | df
def normalize(group):
return (group - group.mean()) / group.std()
df.groupby('key').transform(normalize) | 04 - Groupby operations.ipynb | dpshelio/2015-EuroScipy-pandas-tutorial | bsd-2-clause |
<div class="alert alert-success">
<b>EXERCISE</b>: Calculate the ratio of number roles of actors and actresses to the total number of roles per decade and plot this for both in time (tip: you need to do a groupby twice in two steps, once calculating the numbers, and then the ratios.
</div>
Value counts
A useful shortcut to calculate the number of occurences of certain values is value_counts (this is somewhat equivalent to df.groupby(key).size()))
For example, what are the most occuring movie titles? | titles.title.value_counts().head() | 04 - Groupby operations.ipynb | dpshelio/2015-EuroScipy-pandas-tutorial | bsd-2-clause |
You can skip the following steps if you just want to load the word2vec model I provided... but this is the raw data approach. | # We need to unzip the data file to use it:
!gunzip ../data/yelp/yelp_academic_dataset_reviews.json.gz
# Make sure it is there and unzipped:
!ls -al ../data/yelp/
## Make sure this dataset is here and unzipped.
data = []
with open("../data/yelp/yelp_academic_dataset_reviews.json") as handle:
for line in handle.readlines():
yelp = json.loads(line)
data.append(yelp)
len(data)
data[0]
revs = [d[u'text'] for d in data]
revs[0] | python/Word2Vec_Yelp.ipynb | arnicas/eyeo_nlp | cc0-1.0 |
What is Word2Vec?
"Generally, word2vec is trained using something called a skip-gram model. The skip-gram model, pictures above, attempts to use the vector representation that it learns to predict the words that appear around a given word in the corpus. Essentially, it uses the context of the word as it is used in a variety of books and other literature to derive a meaningful set of numbers. If the “context” of two words is similar, they will have similar vector representations." (Source)
"In word2vec, a distributed representation of a word is used. Take a vector with several hundred dimensions (say 1000). Each word is representated by a distribution of weights across those elements. So instead of a one-to-one mapping between an element in the vector and a word, the representation of a word is spread across all of the elements in the vector, and each element in the vector contributes to the definition of many words.
If I label the dimensions in a hypothetical word vector (there are no such pre-assigned labels in the algorithm of course), it might look a bit like this:"
<img src="img/word2vec-distributed-representation.png">
Source
So that means we can do associative logic, or analogies, with these models:
<img src="img/word2vec-king-queen-vectors.png">
Specifically, a large enough model of the right kind of language (like a lot of news, or lots of books) will allow you to get "queen" from putting in man, king, woman... and doing vector math on them. So, king-man+woman=queen.
Source
<img src="img/word2vec-king-queen-composition.png">
Source
Creating a word2vec Model with Gensim
This takes a while. You don't need to do this, since I already did it. You can skip down to the place where we load the file! | """ An alternate from gensim tutorials - just use all words in the model in a rewiew. No nltk used to split."""
import re
class YelpReviews(object):
"""Iterate over sentences of all plaintext files in a directory """
SPLIT_SENTENCES = re.compile(u"[.!?:]\s+") # split sentences on these characters
def __init__(self, objs, field):
self.field = field
self.objs = objs
def __iter__(self):
for obj in self.objs:
text = obj[self.field]
for sentence in self.SPLIT_SENTENCES.split(text):
yield gensim.utils.simple_preprocess(sentence, deacc=True)
## Don't do this is you already have the model file! Skip to the step after.
## Otherwise, feel free to do it from scratch.
## We pass in the full data objs and use the YelpReviews class to get the 'text' field for us.
#model = gensim.models.Word2Vec(YelpReviews(data, 'text'), min_count=2, workers=2)
#model.save('yelp_w2v_model.mod')
#model.save_word2vec_format('yelp_w2vformat.mod')
# If you already have a model file, load it here:
model = gensim.models.Word2Vec.load_word2vec_format('../data/yelp/yelp_w2vformat.mod')
model.most_similar(positive=["chicken", "waffles"], topn=20)
model.most_similar("waitress")
model.vocab.items()[0:5]
model.most_similar(['good', 'pizza'])
model.most_similar_cosmul(['good', 'pizza']) # less susceptible to extreme outliers
model.most_similar(['dog'])
model.most_similar(['salon'])
model.most_similar(positive=['donuts', 'nypd'], negative=['fireman']) | python/Word2Vec_Yelp.ipynb | arnicas/eyeo_nlp | cc0-1.0 |
Now let's do some basic word sentiment stuff again... for the html side! | import nltk
nltk.data.path = ['../nltk_data']
from nltk.corpus import stopwords
english_stops = stopwords.words('english')
revs[0]
tokens = [nltk.word_tokenize(rev) for rev in revs] # this takes a long time. don't run unless you're sure.
mystops = english_stops + [u"n't", u'...', u"'ve"]
def clean_tokens(tokens, stoplist):
""" Lowercases, takes out punct and stopwords and short strings """
return [token.lower() for token in tokens if (token not in string.punctuation) and
(token.lower() not in stoplist) and len(token) > 2]
clean = [clean_tokens(tok, mystops) for tok in tokens]
from nltk import Text
allclean = [y for x in clean for y in x] # flatten the list of lists
cleantext = Text(allclean)
mostcommon = cleantext.vocab().most_common()[0:1500]
mostcommon_words = [word[0] for word in mostcommon]
mostcommon_words[0:12]
# thing required to get the vectors for tsne
def get_vectors(words, model):
# requires model be in the binary format, not gensim's
word_vectors = []
word_labels = []
for word in words:
if word in model:
word_vectors.append( model[word] )
word_labels.append(word)
return word_vectors, word_labels
mymodel = gensim.models.Word2Vec.load_word2vec_format('../data/yelp/yelp_w2vformat.mod')
vectors, labels = get_vectors(mostcommon_words, mymodel)
# should be same as top words above
labels[:12]
res = ts.tsne(np.asfarray(vectors, dtype='float'), 2, 50, 20) | python/Word2Vec_Yelp.ipynb | arnicas/eyeo_nlp | cc0-1.0 |
The "AFINN-111.txt" file is another sentiment file. | from collections import defaultdict
sentiment = defaultdict(int)
with open('../data/sentiment_wordlists/AFINN-111.txt') as handle:
for line in handle.readlines():
word = line.split('\t')[0]
polarity = line.split('\t')[1]
sentiment[word] = int(polarity)
sentiment['pho']
sentiment['good']
sentiment['angry']
sentiment['pizza']
def render_json( vectors, labels, filename ):
output = []
vectors = np.array(vectors)
for i in range(len(vectors)):
new_hash = {}
new_hash["word"] = str(labels[i])
new_hash["x"] = int(vectors[i][0])
new_hash["y"] = int(vectors[i][1])
new_hash["sentiment"] = sentiment[str(labels[i])]
output.append(new_hash)
with open(filename, 'w') as handle:
json.dump(output, handle)
render_json(res, labels, "../outputdata/yelp.json") | python/Word2Vec_Yelp.ipynb | arnicas/eyeo_nlp | cc0-1.0 |
Import statements | import pandas
import numpy as np
import pyprind | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
GRASS import statements | import grass.script as gscript
from grass.pygrass.vector import VectorTopo
from grass.pygrass.vector.table import DBlinks | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
Function declarations
connect to an attribute table | def connectToAttributeTable(map):
vector = VectorTopo(map)
vector.open(mode='r')
dblinks = DBlinks(vector.c_mapinfo)
link = dblinks[0]
return link.table() | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
finds the nearest element in a vector map (to) for elements in another vector map (from) <br />
calls the GRASS v.distance command | def computeDistance(from_map, to_map):
upload = 'dist'
result = gscript.read_command('v.distance',
from_=from_map,
to=to_map,
upload=upload,
separator='comma',
flags='p')
return result.split('\n') | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
selects vector features from an existing vector map and creates a new vector map containing only the selected features <br />
calls the GRASS v.extract command | def extractFeatures(input_, type_, output):
where = "{0} = '{1}'".format(road_type_field, type_)
gscript.read_command('v.extract',
input_=input_,
where=where,
output=output,
overwrite=True) | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
Get unique 'roads' types | roads_table = connectToAttributeTable(map=roads)
roads_table.filters.select(road_type_field)
cursor = roads_table.execute()
result = np.array(cursor.fetchall())
cursor.close()
road_types = np.unique(result)
print(road_types) | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
Get 'points' attribute table | point_table = connectToAttributeTable(map=points)
point_table.filters.select()
columns = point_table.columns.names()
cursor = point_table.execute()
result = np.array(cursor.fetchall())
cursor.close()
point_data = pandas.DataFrame(result, columns=columns).set_index('cat') | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
Loop through 'roads' types and compute the distances from all 'points' | distances = pandas.DataFrame(columns=road_types, index=point_data.index)
progress_bar = pyprind.ProgBar(road_types.size, bar_char='█', title='Progress', monitor=True, stream=1, width=50)
for type_ in road_types:
# update progress bar
progress_bar.update(item_id=type_)
# extract road data based on type query
extractFeatures(input_=roads, type_=type_, output='roads_tmp')
# compute distance from points to road type
results = computeDistance(points, 'roads_tmp')
# save results to data frame
distances[type_] = [ d.split(',')[1] for d in results[1:len(results)-1] ]
# match index with SiteID
distances['SiteID'] = point_data['ID']
distances.set_index('SiteID', inplace=True) | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
Export distances table to a csv file | distances.to_csv(distance_table_filename, header=False) | Compute distance to roads.ipynb | jacobdein/alpine-soundscapes | mit |
A generalization using accumulation | mapped = list(accumulate(mapped, accumulating))
mapped
clear_cache()
m,v,r = to_matrix_notation(mapped, f, [n-k for k in range(-2, 19)])
m,v,r
m_sym = m.subs(inverted_fibs, simultaneous=True)
m_sym[:,0] = m_sym[:,0].subs(f[2],f[1])
m_sym[1,2] = m_sym[1,2].subs(f[2],f[1])
m_sym
# the following cell produces an error due to ordering, while `m * v` doesn't.
#clear_cache()
#m_sym * v
to_matrix_notation(mapped, f, [n+k for k in range(-18, 3)]) | notebooks/recurrences-unfolding.ipynb | massimo-nocentini/PhD | apache-2.0 |
According to A162741, we can generalize the pattern above: | i = symbols('i')
d = IndexedBase('d')
k_fn_gen = Eq((k+1)*f[n], Sum(d[k,2*k-i]*f[n-i], (i, 0, 2*k)))
d_triangle= {d[0,0]:1, d[n,2*n]:1, d[n,k]:d[n-1, k-1]+d[n-1,k]}
k_fn_gen, d_triangle
mapped = list(accumulate(mapped, accumulating))
mapped
# skip this cell to maintain math coerent version
def adjust(term):
a_wild, b_wild = Wild('a', exclude=[f]), Wild('b')
matched = term.match(a_wild*f[n+2] + b_wild)
return -(matched[a_wild]-1)*f[n+2]
m = fix_combination(mapped,adjust, lambda v, side: Add(v, side))
mapped = list(m)
mapped
to_matrix_notation(mapped, f, [n-k for k in range(-2, 19)])
mapped = list(accumulate(mapped, accumulating))
mapped
to_matrix_notation(mapped, f, [n-k for k in range(-2, 19)])
mapped = list(accumulate(mapped, accumulating))
mapped
to_matrix_notation(mapped, f, [n-k for k in range(-2, 19)])
mapped = list(accumulate(mapped, accumulating))
mapped
to_matrix_notation(mapped, f, [n-k for k in range(-2, 19)]) | notebooks/recurrences-unfolding.ipynb | massimo-nocentini/PhD | apache-2.0 |
Unfolding a recurrence with generic coefficients | s = IndexedBase('s')
a = IndexedBase('a')
swaps_recurrence = Eq(n*s[n],(n+1)*s[n-1]+a[n])
swaps_recurrence
boundary_conditions = {s[0]:Integer(0)}
swaps_recurrence_spec=dict(recurrence_eq=swaps_recurrence, indexed=s,
index=n, terms_cache=boundary_conditions)
unfolded = do_unfolding_steps(swaps_recurrence_spec, 4)
recurrence_eq = project_recurrence_spec(unfolded, recurrence_eq=True)
recurrence_eq
factored_recurrence_eq = project_recurrence_spec(factor_rhs_unfolded_rec(unfolded), recurrence_eq=True)
factored_recurrence_eq
factored_recurrence_eq.rhs.collect(s[n-5]).collect(a[n-4])
factored_recurrence_eq.subs(n,5)
recurrence_eq.subs(n, 5)
def additional_term(n): return (2*Integer(n)-3)/6
as_dict = {a[n]:additional_term(n) for n in range(1,6)}
recurrence_eq.subs(n, 5).subs(as_dict) | notebooks/recurrences-unfolding.ipynb | massimo-nocentini/PhD | apache-2.0 |
A curious relation about Fibonacci numbers, in matrix notation | d = 10
m = Matrix(d,d, lambda i,j: binomial(n-i,j)*binomial(n-j,i))
m
f = IndexedBase('f')
fibs = [fibonacci(i) for i in range(50)]
mp = (ones(1,d)*m*ones(d,1))[0,0]
odd_fibs_eq = Eq(f[2*n+1], mp, evaluate=True)
odd_fibs_eq
(m*ones(d,1)) | notebooks/recurrences-unfolding.ipynb | massimo-nocentini/PhD | apache-2.0 |
Implement Preprocessing Functions
The first thing to do to any dataset is preprocessing. Implement the following preprocessing functions below:
- Lookup Table
- Tokenize Punctuation
Lookup Table
To create a word embedding, you first need to transform the words to ids. In this function, create two dictionaries:
- Dictionary to go from the words to an id, we'll call vocab_to_int
- Dictionary to go from the id to word, we'll call int_to_vocab
Return these dictionaries in the following tuple (vocab_to_int, int_to_vocab) | import numpy as np
import problem_unittests as tests
def create_lookup_tables(text):
"""
Create lookup tables for vocabulary
:param text: The text of tv scripts split into words
:return: A tuple of dicts (vocab_to_int, int_to_vocab)
"""
unique_words = set(text)
vocab_to_int = {word: index for index, word in enumerate(unique_words)}
int_to_vocab = {index: word for index, word in enumerate(unique_words)}
return vocab_to_int, int_to_vocab
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_create_lookup_tables(create_lookup_tables) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Check Point
This is your first checkpoint. If you ever decide to come back to this notebook or have to restart the notebook, you can start from here. The preprocessed data has been saved to disk.
Build the Neural Network
You'll build the components necessary to build a RNN by implementing the following functions below:
- get_inputs
- get_init_cell
- get_embed
- build_rnn
- build_nn
- get_batches
Check the Version of TensorFlow and Access to GPU | """
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper
import numpy as np
import problem_unittests as tests
int_text, vocab_to_int, int_to_vocab, token_dict = helper.load_preprocess()
"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf
# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer'
print('TensorFlow Version: {}'.format(tf.__version__))
# Check for a GPU
if not tf.test.gpu_device_name():
warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
print('Default GPU Device: {}'.format(tf.test.gpu_device_name())) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Input
Implement the get_inputs() function to create TF Placeholders for the Neural Network. It should create the following placeholders:
- Input text placeholder named "input" using the TF Placeholder name parameter.
- Targets placeholder
- Learning Rate placeholder
Return the placeholders in the following tuple (Input, Targets, LearningRate) | def get_inputs():
"""
Create TF Placeholders for input, targets, and learning rate.
:return: Tuple (input, targets, learning rate)
"""
return tf.placeholder(dtype=tf.int32, shape=(None, None), name="input"), tf.placeholder(dtype=tf.int32, shape=(None, None), name="target"), tf.placeholder(dtype=tf.float32, name="learning_rate")
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_inputs(get_inputs) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Build RNN Cell and Initialize
Stack one or more BasicLSTMCells in a MultiRNNCell.
- The Rnn size should be set using rnn_size
- Initalize Cell State using the MultiRNNCell's zero_state() function
- Apply the name "initial_state" to the initial state using tf.identity()
Return the cell and initial state in the following tuple (Cell, InitialState) |
def get_init_cell(batch_size, rnn_size):
"""
Create an RNN Cell and initialize it.
:param batch_size: Size of batches
:param rnn_size: Size of RNNs
:return: Tuple (cell, initialize state)
"""
cell = tf.contrib.rnn.MultiRNNCell([tf.contrib.rnn.BasicLSTMCell(rnn_size)])
return cell, tf.identity(cell.zero_state(batch_size, tf.float32), 'initial_state')
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_init_cell(get_init_cell) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Word Embedding
Apply embedding to input_data using TensorFlow. Return the embedded sequence. | def get_embed(input_data, vocab_size, embed_dim):
"""
Create embedding for <input_data>.
:param input_data: TF placeholder for text input.
:param vocab_size: Number of words in vocabulary.
:param embed_dim: Number of embedding dimensions
:return: Embedded input.
"""
embeddings = tf.Variable(tf.random_uniform((vocab_size, embed_dim), -1, 1))
return tf.nn.embedding_lookup(embeddings, input_data)
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_embed(get_embed) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Build the Neural Network
Apply the functions you implemented above to:
- Apply embedding to input_data using your get_embed(input_data, vocab_size, embed_dim) function.
- Build RNN using cell and your build_rnn(cell, inputs) function.
- Apply a fully connected layer with a linear activation and vocab_size as the number of outputs.
Return the logits and final state in the following tuple (Logits, FinalState) | def build_nn(cell, rnn_size, input_data, vocab_size, embed_dim):
"""
Build part of the neural network
:param cell: RNN cell
:param rnn_size: Size of rnns
:param input_data: Input data
:param vocab_size: Vocabulary size
:param embed_dim: Number of embedding dimensions
:return: Tuple (Logits, FinalState)
"""
embeddings = get_embed(input_data, vocab_size, embed_dim)
outputs, final_state = build_rnn(cell, embeddings) # max_time x batch_size x rnn_size
logits = tf.contrib.layers.fully_connected(outputs, vocab_size, None)
return logits, final_state
# Test was failing because it was written for some previous version of Tensorflow
# tests.test_build_nn(build_nn) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Batches
Implement get_batches to create batches of input and targets using int_text. The batches should be a Numpy array with the shape (number of batches, 2, batch size, sequence length). Each batch contains two elements:
- The first element is a single batch of input with the shape [batch size, sequence length]
- The second element is a single batch of targets with the shape [batch size, sequence length]
If you can't fill the last batch with enough data, drop the last batch.
For exmple, get_batches([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20], 3, 2) would return a Numpy array of the following:
```
[
# First Batch
[
# Batch of Input
[[ 1 2], [ 7 8], [13 14]]
# Batch of targets
[[ 2 3], [ 8 9], [14 15]]
]
# Second Batch
[
# Batch of Input
[[ 3 4], [ 9 10], [15 16]]
# Batch of targets
[[ 4 5], [10 11], [16 17]]
]
# Third Batch
[
# Batch of Input
[[ 5 6], [11 12], [17 18]]
# Batch of targets
[[ 6 7], [12 13], [18 1]]
]
]
```
Notice that the last target value in the last batch is the first input value of the first batch. In this case, 1. This is a common technique used when creating sequence batches, although it is rather unintuitive. | def get_batches(int_text, batch_size, seq_length):
"""
Return batches of input and target
:param int_text: Text with the words replaced by their ids
:param batch_size: The size of batch
:param seq_length: The length of sequence
:return: Batches as a Numpy array
"""
total_batches = len(int_text) // (batch_size * seq_length)
int_text = np.asarray(int_text[:(total_batches * batch_size * seq_length)])
label_text = np.asarray([0] * int_text.shape[0])
label_text[:-1], label_text[-1] = int_text[1:], int_text[0]
int_text = np.reshape(int_text, (-1, batch_size, seq_length))
label_text = np.reshape(label_text, (-1, batch_size, seq_length))
# print(np.concatenate((int_text[:, None, :, :], label_text[:, None, :, :]), axis=1)[0])
return np.concatenate((int_text[:, None, :, :], label_text[:, None, :, :]), axis=1)
# Test was not generic enough i.e. it was passing only a specific arranegement of numbers.
# tests.test_get_batches(get_batches) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Neural Network Training
Hyperparameters
Tune the following parameters:
Set num_epochs to the number of epochs.
Set batch_size to the batch size.
Set rnn_size to the size of the RNNs.
Set embed_dim to the size of the embedding.
Set seq_length to the length of sequence.
Set learning_rate to the learning rate.
Set show_every_n_batches to the number of batches the neural network should print progress. | # Number of Epochs
num_epochs = 100
# Batch Size
batch_size = 256
# RNN Size
rnn_size = 256
# Embedding Dimension Size
embed_dim = 300
# Sequence Length
seq_length = 10
# Learning Rate
learning_rate = 0.01
# Show stats for every n number of batches
show_every_n_batches = 20
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
save_dir = './save' | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Implement Generate Functions
Get Tensors
Get tensors from loaded_graph using the function get_tensor_by_name(). Get the tensors using the following names:
- "input:0"
- "initial_state:0"
- "final_state:0"
- "probs:0"
Return the tensors in the following tuple (InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor) | def get_tensors(loaded_graph):
"""
Get input, initial state, final state, and probabilities tensor from <loaded_graph>
:param loaded_graph: TensorFlow graph loaded from file
:return: Tuple (InputTensor, InitialStateTensor, FinalStateTensor, ProbsTensor)
"""
return (loaded_graph.get_tensor_by_name('input:0'), loaded_graph.get_tensor_by_name('initial_state:0'),
loaded_graph.get_tensor_by_name('final_state:0'), loaded_graph.get_tensor_by_name('probs:0'))
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_get_tensors(get_tensors) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Choose Word
Implement the pick_word() function to select the next word using probabilities. | def pick_word(probabilities, int_to_vocab):
"""
Pick the next word in the generated text
:param probabilities: Probabilites of the next word
:param int_to_vocab: Dictionary of word ids as the keys and words as the values
:return: String of the predicted word
"""
probabilities = np.reshape(probabilities, (len(int_to_vocab.keys()),))
return int_to_vocab[np.argmax(probabilities)]
"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_pick_word(pick_word) | tv-script-generation/dlnd_tv_script_generation.ipynb | mu4farooqi/deep-learning-projects | gpl-3.0 |
Index comparison
See how indexes affect queries
First without then with |
scratch_db.zips.drop_indexes()
count = scratch_db.zips.find().count()
city_count = scratch_db.zips.find({"city": "FLAGSTAFF"}).count()
city_explain = scratch_db.zips.find({"city": "FLAGSTAFF"}).explain()['executionStats']
print(count)
print(city_count)
print(city_explain)
scratch_db.zips.drop_indexes()
scratch_db.zips.create_index([("city", ASCENDING)])
count = scratch_db.zips.find().count()
city_count = scratch_db.zips.find({"city": "FLAGSTAFF"}).count()
city_explain = scratch_db.zips.find({"city": "FLAGSTAFF"}).explain()['executionStats']
print(count)
print(city_count)
print(city_explain) | MongoDB.ipynb | 0Rick0/Fontys-DS-GCD | mit |
You can see with the index it's execution is a bit different.
Seeing the executionTimeMillis parameter shows that the second one is executed much faster.
This is because the index allow you to search the index instead of all the documents.
Some other information about the dataset | print("Amount of cities per state:")
pipeline = [
{"$unwind": "$state"},
{"$group": {"_id": "$state", "count": {"$sum": 1}}},
{"$sort": SON([("count", -1), ("_id", -1)])}
]
results = scratch_db.zips.aggregate(pipeline)
for result in results:
print("State %s: %d" % tuple(result.values()))
print("Amount of cities with fewer then 50 people")
lt = scratch_db.zips.find({"pop": {"$lt": 50}})
print("%d cities" % lt.count())
for city in lt.limit(10):
print("%s: %d" % (city['city'], city['pop'])) | MongoDB.ipynb | 0Rick0/Fontys-DS-GCD | mit |
Geolocation
Mongodb also has build in support for geolocation indexes
This allows for searching for example nearby shops for a given location | scratch_db.zips.create_index([("loc", "2dsphere")])
flagstaff = scratch_db.zips.find_one({"city": "FLAGSTAFF"})
nearby = scratch_db.zips.find({"loc": {
"$near": {
"$geometry": {
'type': 'Point',
'coordinates': flagstaff['loc']
},
"$maxDistance": 50000
}
}})
for city in nearby:
print(city['city']) | MongoDB.ipynb | 0Rick0/Fontys-DS-GCD | mit |
Create a managed tabular dataset from a CSV
A Managed dataset can be used to create an AutoML model or a custom model. | # Create a managed tabular dataset
ds = # TODO 1: Your code goes here(display_name="abalone", gcs_source=[gcs_csv_path])
ds.resource_name | courses/machine_learning/deepdive2/production_ml/labs/sdk_metric_parameter_tracking_for_custom_jobs.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
Start a new experiment run to track training parameters and start the training job. Note that this operation will take around 10 mins. | aiplatform.start_run("custom-training-run-1") # Change this to your desired run name
parameters = {"epochs": 10, "num_units": 64}
aiplatform.log_params(parameters)
# Launch the training job
model = # TODO 2: Your code goes here(
ds,
replica_count=1,
model_display_name="abalone-model",
args=[f"--epochs={parameters['epochs']}", f"--num_units={parameters['num_units']}"],
) | courses/machine_learning/deepdive2/production_ml/labs/sdk_metric_parameter_tracking_for_custom_jobs.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
Deploy Model and calculate prediction metrics
Deploy model to Google Cloud. This operation will take 10-20 mins. | # Deploy the model
endpoint = # TODO 3: Your code goes here(machine_type="n1-standard-4") | courses/machine_learning/deepdive2/production_ml/labs/sdk_metric_parameter_tracking_for_custom_jobs.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
Perform online prediction. | # Perform online prediction using endpoint
prediction = # TODO 4: Your code goes here(test_dataset.tolist())
prediction | courses/machine_learning/deepdive2/production_ml/labs/sdk_metric_parameter_tracking_for_custom_jobs.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
Extract all parameters and metrics created during this experiment. | # Extract all parameters and metrics of the experiment
# TODO 5: Your code goes here | courses/machine_learning/deepdive2/production_ml/labs/sdk_metric_parameter_tracking_for_custom_jobs.ipynb | GoogleCloudPlatform/training-data-analyst | apache-2.0 |
XGBoost HP Tuning on AI Platform
This notebook trains a model on Ai Platform using Hyperparameter Tuning to predict a car's Miles Per Gallon. It uses Auto MPG Data Set from UCI Machine Learning Repository.
Citation: Dua, D. and Karra Taniskidou, E. (2017). UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. Irvine, CA: University of California, School of Information and Computer Science.
How to train your model on AI Platform with HP tuning.
Using HP Tuning for training can be done in a few steps:
1. Create your python model file
1. Add argument parsing for the hyperparameter values. (These values are chosen for you in this notebook)
1. Add code to download your data from Google Cloud Storage so that AI Platform can use it
1. Add code to track the performance of your hyperparameter values.
1. Add code to export and save the model to Google Cloud Storage once AI Platform finishes training the model
1. Prepare a package
1. Submit the training job
Prerequisites
Before you jump in, let’s cover some of the different tools you’ll be using to get HP tuning up and running on AI Platform.
Google Cloud Platform lets you build and host applications and websites, store data, and analyze data on Google's scalable infrastructure.
AI Platform is a managed service that enables you to easily build machine learning models that work on any type of data, of any size.
Google Cloud Storage (GCS) is a unified object storage for developers and enterprises, from live data serving to data analytics/ML to data archiving.
Cloud SDK is a command line tool which allows you to interact with Google Cloud products. In order to run this notebook, make sure that Cloud SDK is installed in the same environment as your Jupyter kernel.
Overview of Hyperparameter Tuning - Hyperparameter tuning takes advantage of the processing infrastructure of Google Cloud Platform to test different hyperparameter configurations when training your model.
Part 0: Setup
Create a project on GCP
Create a Google Cloud Storage Bucket
Enable AI Platform Training and Prediction and Compute Engine APIs
Install Cloud SDK
Install XGBoost [Optional: used if running locally]
Install pandas [Optional: used if running locally]
Install cloudml-hypertune [Optional: used if running locally]
These variables will be needed for the following steps.
* TRAINER_PACKAGE_PATH <./auto_mpg_hp_tuning> - A packaged training application that will be staged in a Google Cloud Storage location. The model file created below is placed inside this package path.
* MAIN_TRAINER_MODULE <auto_mpg_hp_tuning.train> - Tells AI Platform which file to execute. This is formatted as follows <folder_name.python_file_name>
* JOB_DIR <gs://$BUCKET_ID/xgboost_learn_job_dir> - The path to a Google Cloud Storage location to use for job output.
* RUNTIME_VERSION <1.9> - The version of AI Platform to use for the job. If you don't specify a runtime version, the training service uses the default AI Platform runtime version 1.0. See the list of runtime versions for more information.
* PYTHON_VERSION <3.5> - The Python version to use for the job. Python 3.5 is available with runtime version 1.4 or greater. If you don't specify a Python version, the training service uses Python 2.7.
* HPTUNING_CONFIG <hptuning_config.yaml> - Path to the job configuration file.
Replace:
* PROJECT_ID <YOUR_PROJECT_ID> - with your project's id. Use the PROJECT_ID that matches your Google Cloud Platform project.
* BUCKET_ID <YOUR_BUCKET_ID> - with the bucket id you created above.
* JOB_DIR <gs://YOUR_BUCKET_ID/xgboost_job_dir> - with the bucket id you created above.
* REGION <REGION> - select a region from here or use the default 'us-central1'. The region is where the model will be deployed. | # Replace <PROJECT_ID> and <BUCKET_ID> with proper Project and Bucket ID's:
%env PROJECT_ID <PROJECT_ID>
%env BUCKET_ID <BUCKET_ID>
%env JOB_DIR gs://<BUCKET_ID>/xgboost_job_dir
%env REGION us-central1
%env TRAINER_PACKAGE_PATH ./auto_mpg_hp_tuning
%env MAIN_TRAINER_MODULE auto_mpg_hp_tuning.train
%env RUNTIME_VERSION 1.9
%env PYTHON_VERSION 3.5
%env HPTUNING_CONFIG hptuning_config.yaml
! mkdir auto_mpg_hp_tuning | notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
The data
The Auto MPG Data Set that this sample
uses for training is provided by the UC Irvine Machine Learning
Repository. We have hosted the data on a public GCS bucket gs://cloud-samples-data/ml-engine/auto_mpg/. The data has been pre-processed to remove rows with incomplete data so as not to create additional steps for this notebook.
Training file is auto-mpg.data
Note: Your typical development process with your own data would require you to upload your data to GCS so that AI Platform can access that data. However, in this case, we have put the data on GCS to avoid the steps of having you download the data from UC Irvine and then upload the data to GCS.
Citation: Dua, D. and Karra Taniskidou, E. (2017). UCI Machine Learning Repository [http://archive.ics.uci.edu/ml]. Irvine, CA: University of California, School of Information and Computer Science.
Disclaimer
This dataset is provided by a third party. Google provides no representation,
warranty, or other guarantees about the validity or any other aspects of this dataset.
Part 1: Create your python model file
First, we'll create the python model file (provided below) that we'll upload to AI Platform. This is similar to your normal process for creating a XGBoost model. However, there are a few key differences:
1. Downloading the data from GCS at the start of your file, so that AI Platform can access the data.
1. Exporting/saving the model to GCS at the end of your file, so that you can use it for predictions.
1. Define a command-line argument in your main training module for each tuned hyperparameter.
1. Use the value passed in those arguments to set the corresponding hyperparameter in your application's XGBoost code.
1. Use cloudml-hypertune to track your training jobs metrics.
The code in this file first handles the hyperparameters passed to the file from AI Platform. Then it loads the data into a pandas DataFrame that can be used by XGBoost. Then the model is fit against the training data and the metrics for that data are shared with AI Platform. Lastly, Python's built in pickle library is used to save the model to a file that can be uploaded to AI Platform's prediction service.
Note: In normal practice you would want to test your model locally on a small dataset to ensure that it works, before using it with your larger dataset on AI Platform. This avoids wasted time and costs.
Setup the imports and helper functions | %%writefile ./auto_mpg_hp_tuning/train.py
import argparse
import datetime
import os
import pandas as pd
import subprocess
import pickle
from google.cloud import storage
import hypertune
import xgboost as xgb
from random import shuffle
def split_dataframe(dataframe, rate=0.8):
indices = dataframe.index.values.tolist()
length = len(dataframe)
shuffle(indices)
train_size = int(length * rate)
train_indices = indices[:train_size]
test_indices = indices[train_size:]
return dataframe.iloc[train_indices], dataframe.iloc[test_indices]
| notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
Load the hyperparameter values that are passed to the model during training.
In this tutorial, the Lasso regressor is used, because it has several parameters that can be used to help demonstrate how to choose HP tuning values. (The range of values are set below in the configuration file for the HP tuning values.) | %%writefile -a ./auto_mpg_hp_tuning/train.py
parser = argparse.ArgumentParser()
parser.add_argument(
'--job-dir', # handled automatically by AI Platform
help='GCS location to write checkpoints and export models',
required=True
)
parser.add_argument(
'--max_depth', # Specified in the config file
help='Maximum depth of the XGBoost tree. default: 3',
default=3,
type=int
)
parser.add_argument(
'--n_estimators', # Specified in the config file
help='Number of estimators to be created. default: 100',
default=100,
type=int
)
parser.add_argument(
'--booster', # Specified in the config file
help='which booster to use: gbtree, gblinear or dart. default: gbtree',
default='gbtree',
type=str
)
args = parser.parse_args()
| notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
Add code to download the data from GCS
In this case, using the publicly hosted data,AI Platform will then be able to use the data when training your model. | %%writefile -a ./auto_mpg_hp_tuning/train.py
# Public bucket holding the auto mpg data
bucket = storage.Client().bucket('cloud-samples-data')
# Path to the data inside the public bucket
blob = bucket.blob('ml-engine/auto_mpg/auto-mpg.data')
# Download the data
blob.download_to_filename('auto-mpg.data')
# ---------------------------------------
# This is where your model code would go. Below is an example model using the auto mpg dataset.
# ---------------------------------------
# Define the format of your input data including unused columns
# (These are the columns from the auto-mpg data files)
COLUMNS = [
'mpg',
'cylinders',
'displacement',
'horsepower',
'weight',
'acceleration',
'model-year',
'origin',
'car-name'
]
FEATURES = [
'cylinders',
'displacement',
'horsepower',
'weight',
'acceleration',
'model-year',
'origin'
]
TARGET = 'mpg'
# Load the training auto mpg dataset
with open('./auto-mpg.data', 'r') as train_data:
raw_training_data = pd.read_csv(train_data, header=None, names=COLUMNS, delim_whitespace=True)
raw_training_data = raw_training_data[FEATURES + [TARGET]]
train_df, test_df = split_dataframe(raw_training_data, 0.8)
| notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
Use the Hyperparameters
Use the Hyperparameter values passed in those arguments to set the corresponding hyperparameters in your application's XGBoost code. | %%writefile -a ./auto_mpg_hp_tuning/train.py
# Create the regressor, here we will use a Lasso Regressor to demonstrate the use of HP Tuning.
# Here is where we set the variables used during HP Tuning from
# the parameters passed into the python script
regressor = xgb.XGBRegressor(max_depth=args.max_depth,
n_estimators=args.n_estimators,
booster=args.booster
)
# Transform the features and fit them to the regressor
regressor.fit(train_df[FEATURES], train_df[TARGET])
| notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
Report the mean accuracy as hyperparameter tuning objective metric. | %%writefile -a ./auto_mpg_hp_tuning/train.py
# Calculate the mean accuracy on the given test data and labels.
score = regressor.score(test_df[FEATURES], test_df[TARGET])
# The default name of the metric is training/hptuning/metric.
# We recommend that you assign a custom name. The only functional difference is that
# if you use a custom name, you must set the hyperparameterMetricTag value in the
# HyperparameterSpec object in your job request to match your chosen name.
# https://cloud.google.com/ml-engine/reference/rest/v1/projects.jobs#HyperparameterSpec
hpt = hypertune.HyperTune()
hpt.report_hyperparameter_tuning_metric(
hyperparameter_metric_tag='my_metric_tag',
metric_value=score,
global_step=1000)
| notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
Export and save the model to GCS | %%writefile -a ./auto_mpg_hp_tuning/train.py
# Export the model to a file
model_filename = 'model.pkl'
with open(model_filename, "wb") as f:
pickle.dump(regressor, f)
# Example: job_dir = 'gs://BUCKET_ID/xgboost_job_dir/1'
job_dir = args.job_dir.replace('gs://', '') # Remove the 'gs://'
# Get the Bucket Id
bucket_id = job_dir.split('/')[0]
# Get the path
bucket_path = job_dir[len('{}/'.format(bucket_id)):] # Example: 'xgboost_job_dir/1'
# Upload the model to GCS
bucket = storage.Client().bucket(bucket_id)
blob = bucket.blob('{}/{}'.format(
bucket_path,
model_filename))
blob.upload_from_filename(model_filename)
| notebooks/xgboost/HyperparameterTuningWithXGBoostInCMLE.ipynb | GoogleCloudPlatform/cloudml-samples | apache-2.0 |
Subsets and Splits