RazinAleks/SO-Python_QA-Data_Science_and_Machine_Learning_class

0

43,217,958

0

1

false

24

2017-04-04T21:22:00.000

8

3

0

Pandas data precision

43,217,916

1

python,pandas

Your data is stored with the precision, corresponding to your dtype (np.float16, np.float32, np.float64). pd.options.display.precision - allows you to change the precision for printing the data

By default the numerical values in data frame are stored up to 6 decimals only. How do I get the full precision. For example 34.98774564765 is stored as 34.987746. I do want the full value. and 0.00000565 is stored as 0. . Apart from applying formats to each data frame is there any global setting that helps preserving the precision. Thanks

0

1

80,695

0

43,219,694

0

1

false

0

2017-04-04T23:54:00.000

0

2

0

How to Omit NaN values when applying groupyby in Pandas

43,219,641

0

python,pandas

I get that the mean of that particular group is NAN when a NAN value is present FALSE! :) the mean will only consider non null values. You are safe my man.

I have a dataset consisting of multiple columns and I want to calculate the average by using the groupby function in Python. However, since some of the values are NAN I get that the mean of that particular group is NAN when a NAN value is present. I would like to omit this value, not set it to zero or fill it with any statistical variable, just omit. Any idea how I can achieve this? Thanks in advance!

0

1

266

0

53,868,885

0

1

0

2

false

6

2017-04-04T23:58:00.000

1

5

1

Segmentation fault when I try to run Anaconda Navigator

43,219,679

0.039979

python,ubuntu,anaconda,navigator

I had the same issue when I install OpenCV library using conda.Most probably downgrading something makes this issue happen. Just type : conda update --all

I have recently Installed Anaconda for Python 3.6 but it shows the error "Segmentation fault" whenever I try to run Anaconda-Navigator. I've tried just writting in the terminal Anaconda-Navigator and also going to my Anaconda3 folder and try to execute it inside bin. The only solution that works so far is accessing the previously bin folder as root. My problem is that I need to activate TensorFlow before I run anything in my console, but that is imposible as a root user. I've already try to upgrade both, Anaconda and Navigator and reinstall them but nothing ocurrs Anyone here has any idea of what is happening?

0

1

9,783

0

47,718,983

0

1

0

2

false

6

2017-04-04T23:58:00.000

0

5

1

Segmentation fault when I try to run Anaconda Navigator

43,219,679

0

python,ubuntu,anaconda,navigator

I had the same problem.I solved it by adding /lib to mt LD_LIBRARY_PATH. Note: On my system Anaconda installation path is /home/pushyamik/anaconda3.

I have recently Installed Anaconda for Python 3.6 but it shows the error "Segmentation fault" whenever I try to run Anaconda-Navigator. I've tried just writting in the terminal Anaconda-Navigator and also going to my Anaconda3 folder and try to execute it inside bin. The only solution that works so far is accessing the previously bin folder as root. My problem is that I need to activate TensorFlow before I run anything in my console, but that is imposible as a root user. I've already try to upgrade both, Anaconda and Navigator and reinstall them but nothing ocurrs Anyone here has any idea of what is happening?

0

1

9,783

0

43,304,907

0

1

false

0

2017-04-06T16:32:00.000

0

2

0

Generating an SIS epidemilogical model using Python networkx

43,260,916

0

python,graph,graph-theory,networkx

Do you use networkx for calculation or visualization? There is no need to use it for calculation since your model is simple and it is easier to calculate it with matrix (vector) operations. That is suitable for numpy. Main part in a step is calculation of probability of switching from 0 to 1. Let N be vector that for each node stores 0 or 1 depending of state. Than probability that node n switch from 0 to 1 is numpy.amax(A[n,:] * N). If you need visualization, than probably there are better libraries than networkx.

I have been told networkx library in python is the standard library to use for graph-theoretical applications, but I have found using it quite frustrating so far. What I want to do is this: Generating an SIS epidemiological network, assigning initial contact rates and recovery rates and then following the progress of the disease. More precisely, imagine a network of n individuals and an adjacency matrix A. Values of A are in [0,1] range and are contact rates. This means that the (i,j) entry shows the probability that disease is transferred from node i to node j. Initially, each node is assigned a random label, which can be either 1 (for Infective individuals) or 0 (for Susceptible, which are the ones which have not caught the disease yet). At each time-step, if the node has a 0 label, then with a probability equal to the maximum value of weights for incoming edges to the node, it can turn into a 1. If the node has a 1 label then with a probability specified by its recovery rate, it can turn into a 0. Recovery rate is a value assigned to each node at the beginning of the simulation, and is in [0,1] range. And while the network evolves in each time step, I want to display the network with each node label coloured differently. If somebody knows of any other library in python that can do such a thing more efficiently than netwrokx, be grateful if you let me know.

0

1

1,096

0

65,609,395

0

1

false

0

2017-04-06T20:14:00.000

0

3

0

Is there a way to do inverse z-transforms in Python?

43,264,959

0

python,inverse-transform

You can use scipy.signal.residuez for z^-n form of z-transform or scipy.signal.residue for z^n form of z-transform.

Is there a way to do inverse z-transforms in Python? (I don’t see anything like this in NumPy or SciPy).

0

1

4,204

0

43,362,812

0

1

true

0

2017-04-07T10:10:00.000

1

2

0

Parallel training with CNTK and numpy interop

43,275,431

1.2

python,cntk

There are the following options: 1) Use distributed learner + training session - then you need to either use ImageDeserializer, or implement your own MinibatchSource (this extensibility only available starting RC2) 2) Use distributed learner + write the training loop yourself. In that case you have to take care of splitting the data (each worker should only read images that correspond to its rank) and all conditions inside the loop should be based on trainer->TotalNumberOfSamples() (i.e. checkpointing if you do any).

I'm training an autoencoder network which needs to read in three images per training sample (one input RGB image, two output RGB images). It was easy to make this work with python and numpy interop and reading the image files in myself. How can I enable parallel/distributed training with this? Do I have to use the training session construct? Do I have to use the image reader minibatch source with that?

0

1

194

0

43,288,022

0

1

false

0

2017-04-07T22:31:00.000

1

0

Pythonic array indexing with boolean masking array

43,287,990

0.197375

python,numpy

Negating a boolean mask array in NumPy is ~mask. Also, consider whether you actually need where at all. Seemingly the most common use is some_array[np.where(some_mask)], but that's just an unnecessarily wordy and inefficient way to write some_array[some_mask].

I get a PEP8 complaint about numpy.where(mask == False) where mask is a boolean array. The PEP8 recommendation comparison should be either 'if condition is false' or 'if not condition'. What is the pythonic syntax for the suggested comparison inside numpy.where()?

0

1

905

0

43,311,283

0

1

false

3

2017-04-09T18:51:00.000

0

2

0

Spark and Cassandra through Python

43,310,597

0

python,apache-spark,cassandra,pyspark

I'll just give my "short" 2 cents. The official docs are totally fine for you to get started. You might want to specify why this isn't working, i.e. did you run out of memory (perhaps you just need to increase the "driver" memory) or is there some specific error that is causing your example not to work. Also it would be nice if you provided that example. Here are some of my opinions/experiences that I had. Usually, not always, but most of the time you have multiple columns in partitions. You don't always have to load all the data in a table and more or less you can keep the processing (most of the time) within a single partition. Since the data is sorted within a partition this usually goes pretty fast. And didn't present any significant problem. If you don't want the whole store in casssandra fetch to spark cycle to do your processing you have really a lot of the solutions out there. Basically that would be quora material. Here are some of the more common one: Do the processing in your application right away - might require some sort of inter instance communication framework like hazelcast of even better akka cluster this is really a wide topic spark streaming - just do your processing right away in micro batching and flush results for reading to some persistence layer - might be cassandra apache flink - use proper streaming solution and periodically flush state of the process to i.e. cassandra Store data into cassandra the way it's supposed to be read - this approach is the most adviseable (just hard to say with the info you provided) The list could go on and on ... User defined function in cassandra, aggregate functions if your task is something simpler. It might be also a good idea that you provide some details about your use case. More or less what I said here is pretty general and vague, but then again putting this all into a comment just wouldn't make sense.

I have huge data stored in cassandra and I wanted to process it using spark through python. I just wanted to know how to interconnect spark and cassandra through python. I have seen people using sc.cassandraTable but it isnt working and fetching all the data at once from cassandra and then feeding to spark doesnt make sense. Any suggestions?

0

1

1,860

0

45,893,216

0

2

false

1

2017-04-10T07:25:00.000

0

3

0

Python ImportError - NUMPY MKL

43,317,119

0

python,python-import,importerror

Actually, there is no problem. Just need to restart the Jupyter and you will see it is working well.

I keep getting this error whenever I try running this code, if I can get some insight in what is going on that would be a great help, since I'm pretty new to this coding environment I would really appreciate some help. The code is this: File "C:\Users\user\Desktop\Python\pythonsimulation.py", line 6, in from scipy import * File "C:\Python34\lib\site-packages\scipy__init__.py", line 61, in from numpy._distributor_init import NUMPY_MKL # requires numpy+mkl ImportError: cannot import name 'NUMPY_MKL' Am I missing some package or module that python requires in order to run the code?

0

1

5,133

0

45,571,159

0

2

false

1

2017-04-10T07:25:00.000

0

3

0

Python ImportError - NUMPY MKL

43,317,119

0

python,python-import,importerror

If you are using Jupyter, try restarting the kernel. please click Restart in Kernel menu.

I keep getting this error whenever I try running this code, if I can get some insight in what is going on that would be a great help, since I'm pretty new to this coding environment I would really appreciate some help. The code is this: File "C:\Users\user\Desktop\Python\pythonsimulation.py", line 6, in from scipy import * File "C:\Python34\lib\site-packages\scipy__init__.py", line 61, in from numpy._distributor_init import NUMPY_MKL # requires numpy+mkl ImportError: cannot import name 'NUMPY_MKL' Am I missing some package or module that python requires in order to run the code?

0

1

5,133

0

50,902,599

0

1

false

3

2017-04-10T13:39:00.000

2

4

0

How to simply kill python-tensorflow process and release memory?

43,324,788

0.099668

python,tensorflow

Don't run this on your desktop, but for HPC/remote machines with no display, this kills all left over GPU-using processes: nvidia-smi -q -d PIDS | grep -P "Process ID +: [0-9]+" | grep -Po "[0-9]+" | xargs kill -9

Whenever I run a python script that uses tensorflow and for some reason decide to kill it before it finishes, there is the problem that ctrl-c doesn't work. I would use ctrl-z but it doesn't release the gpu memory, so when i try to re-run the script there is no memory left. Is there a solution for this in linux?

0

1

6,349

0

43,386,604

0

1

false

3

2017-04-10T19:29:00.000

1

2

0

How to train an SVM classifier on a satellite image using Python

43,331,510

0.099668

python,machine-learning,scikit-learn,svm,k-means

My Solution:- Manual Processing:- If the size of your dataset is small, you can manually create a vector data (also reliable, when it is created by yourself). If not, it is much difficult to apply SVM to classify the images. Automatic Processing:- Step 1:- You can use "Unsupervised Image Clustering" technique to group your images into those 4 categories, then label the images from 1 to 4 after clustering is done. (eg. K-Means Clustering Algorithm) Step 2:- Currently, you are having a dataset of labeled images. Split them to train-test data. Step 3:- Now apply SVM to classify your test images and find out your model accuracy.

I am using scikit-learn library to perform a supervised classification (Support Vector Machine classifier) on a satellite image. My main issue is how to train my SVM classifier. I have watched many videos on youtube and have read a few tutorials on how to train an SVM model in scikit-learn. All the tutorials I have watched, they used the famous Iris datasets. In order to perform a supervised SVM classification in scikit-learn we need to have labels. For Iris datasets we have the Iris.target which is the labels ('setosa', 'versicolor', 'virginica') we are trying to predict. The procedure of training is straightforward by reading the scikit-learn documentation. In my case, I have to train a SAR satellite image captured over an urban area and I need to classify the urban area, roads, river and vegetation (4 classes). This image has two bands but I do not have label data for each class I am trying to predict such as the Iris data. So, my question is, do I have to manually create vector data (for the 4 classes) in order to train the SVM model? Is there an easier way to train the model than manually creating vector data? What do we do in this case? I am bit confused to be honest. I would appreciate any help

0

1

6,775

0

43,337,961

0

1

0

2

false

0

2017-04-11T05:50:00.000

0

4

0

python : create a variable with different dimension sizes

43,337,694

0

python,variables,dimensions

Is not possible to make arrays with different sizes as I understood you want to, and this is because a 2D-array is basically a table with rows and columns, and each row has the same number of columns, no matter what. But, you can join the values in each variable and save the resulting strings in the array, and to use them again just split it back and parse the values to the type you need them.

I want to create a variable D by combining two other variable x and y. x has the shape [731] and y has the shape [146]. At the end D should be 2D so that D[0] contains all x-values and D[1] all y-values. I hope I explained it in a way someone can understand what I want to do. Can someone help me with this?

0

1

792

0

43,337,785

0

1

0

2

true

0

2017-04-11T05:50:00.000

2

4

0

python : create a variable with different dimension sizes

43,337,694

1.2

python,variables,dimensions

It is a simple as: D = [x, y] Hope it helped :)

I want to create a variable D by combining two other variable x and y. x has the shape [731] and y has the shape [146]. At the end D should be 2D so that D[0] contains all x-values and D[1] all y-values. I hope I explained it in a way someone can understand what I want to do. Can someone help me with this?

0

1

792

0

45,476,167

0

3

false

8

2017-04-11T09:48:00.000

0

4

0

How Yolo calculate P(Object) in the YOLO 9000

43,342,433

0

python,tensorflow,object-detection,yolo

There are 13x13 grid cells, true, but P(object) is calculated for each of 5x13x13 anchor boxes. From the YOLO9000 paper: When we move to anchor boxes we also decouple the class prediction mechanism from the spatial location and instead predict class and objectness for every anchor box. I can't comment yet because I'm new here, but if you're wondering about test time, it works kind of like an RPN. At each grid cell, the 5 anchor boxes each predict a bounding box, which can be larger than the grid cell, and then non-maximum suppression is used to pick the top few boxes to do classification on. P(object) is just a probability, the network doesn't "know" if there is really an object in there or not. You can also look at the source code for the forward_region_layer method in region_layer.c and trace how the losses are calculated, if you're interested.

Currently I am testing the yolo 9000 model for object detection and in the Paper I understand that the image is splited in 13X13 boxes and in each boxes we calculate P(Object), but How can we calculate that ? how can the model know if there is an object in this boxe or not, please I need help to understand that I am using tensorflow Thanks,

0

1

1,673

0

59,667,414

0

3

false

8

2017-04-11T09:48:00.000

0

4

0

How Yolo calculate P(Object) in the YOLO 9000

43,342,433

0

python,tensorflow,object-detection,yolo

During test time the YOLO network gets the IOU from the default setted value. That is 0.5.

Currently I am testing the yolo 9000 model for object detection and in the Paper I understand that the image is splited in 13X13 boxes and in each boxes we calculate P(Object), but How can we calculate that ? how can the model know if there is an object in this boxe or not, please I need help to understand that I am using tensorflow Thanks,

0

1

1,673

0

44,433,597

0

3

false

8

2017-04-11T09:48:00.000

4

0

How Yolo calculate P(Object) in the YOLO 9000

43,342,433

0.197375

python,tensorflow,object-detection,yolo

They train for the confidence score = P(object) * IOU. For the ground truth box they take P(object)=1 and for rest of the grid pixels the ground truth P(object) is zero. You are training your network to tell you if some object in that grid location i.e. output 0 if not object, output IOU if partial object and output 1 if object is present. So at test time, your model has become capable of telling if there is an object at that location.

Currently I am testing the yolo 9000 model for object detection and in the Paper I understand that the image is splited in 13X13 boxes and in each boxes we calculate P(Object), but How can we calculate that ? how can the model know if there is an object in this boxe or not, please I need help to understand that I am using tensorflow Thanks,

0

1

1,673

0

43,345,587

0

1

true

4

2017-04-11T10:52:00.000

2

1

0

Feature extraction in Keras

43,343,868

1.2

python,image-processing,computer-vision,deep-learning,keras

There is no general rules on how to choose the layer for feature extraction but you might use a easy rule of thumb. The deeper you go to the network - the less ImageNet specific semantic features you would have. But in the same time - you are getting less semantic features also. What I would do is to use the pool layers in both topologies - and if this didn't work well - then I would go deeper by setting the depth as metaparameter.

I am using pretrained resnet50 and inception v3 networks to extract features from my images, which I then use with my ML algo. Which layers are recommended for feature extraction? I am currently using: "mixed10" in Inception v3 and "avg_pool" in resent50. The features are modelling well in XGBoost though. Thank you.

0

1

1,958

0

60,558,628

0

3

false

1

2017-04-11T12:27:00.000

0

3

0

Best Way to create a bounding box for object detection

43,345,925

0

python,tensorflow,deep-learning,object-detection

There is not any rule as such but yes the best practice of proper annotation is to keep certain pixel width while creating bounding boxes. See, the background changes are the variations in the object which will make it robust but keep in mind to have enough samples to properly recognizing the patterns in object(edges, shapes, textures, etc.) Hope i addressed your query!

Currently, I am working to create a deep neural network for object detection, and i am also create my own dataset, and I use the bounding box to annotate my images, and my question is what are the rules to have the best bounding box for my images training. I mean if I wrap my object is it good to limit the background of my object or do I need t find a way to bound only my object. Thanks,

0

1

2,423

0

53,543,602

0

3

false

1

2017-04-11T12:27:00.000

0

3

0

Best Way to create a bounding box for object detection

43,345,925

0

python,tensorflow,deep-learning,object-detection

You can reference YOLO algorithm- this is the best algorithm for object detection. The first, input image will divide into SxS grid cell, Yolo will predict 5 bounding box for each cell and with each bounding box, Yolo also predict the center coordinates of box, width, height of box and confidence score of having any object in that box along with the probabilities that object will belong to M classes. After that, we use Non Max Suppression and IOU to calculate the accuracy between bounding box with ground truth and get only the most exactly bounding box for the object in the input image.

Currently, I am working to create a deep neural network for object detection, and i am also create my own dataset, and I use the bounding box to annotate my images, and my question is what are the rules to have the best bounding box for my images training. I mean if I wrap my object is it good to limit the background of my object or do I need t find a way to bound only my object. Thanks,

0

1

2,423

0

43,346,462

0

3

false

1

2017-04-11T12:27:00.000

0

3

0

Best Way to create a bounding box for object detection

43,345,925

0

python,tensorflow,deep-learning,object-detection

I am not specialized in bounding box, but in general in deep learning, we try to obtain a network which will be robust against irrelevant variables, in your case, the background. Bounding should not be dependant of the background, so set your bounding box the way you want, it should be learned by the network how to replicate it. The most important thing is more the size of the database and having consistent bounding than precise bounding. Also if you want your network to be robust against changes of backgrounds, you should have as many backgrounds as possible, uncorrelated with the bounding.

Currently, I am working to create a deep neural network for object detection, and i am also create my own dataset, and I use the bounding box to annotate my images, and my question is what are the rules to have the best bounding box for my images training. I mean if I wrap my object is it good to limit the background of my object or do I need t find a way to bound only my object. Thanks,

0

1

2,423

0

43,352,367

0

1

0

1

false

1

2017-04-11T16:39:00.000

0

2

0

R, Python and pyRserve - multi-threaded examples?

43,351,701

0

python,r,multithreading,rserve,pyrserve

I kept on working with the code and it turns out that each thread needs its own port in order to work. I didn't find that documented anywhere, I was just trying out different idea. So: I set up as many instances of Rserve as a I wanted threads. Each one of those instances was its own port In my python code, when I instantiated the pyRserve object, I assigned it a unique port number. Multi-threading is now working as desired and fast!

I have a Python script set up where it instantiates Rserve, sets up a few R scripts and functions and then runs some data against the functions. However, I have been unable to create a multi-threaded instance of this same process. My core issue is that one thread always seems to dominate the processing and all of the other threads are ignored. I've made the assumption that pyRserve can be multi-threaded - is that a correct assumption? Are there any examples out there that show this as a multi-threaded app?

0

1

290

0

43,369,916

0

1

false

1

2017-04-11T16:42:00.000

2

0

how to import cv2 and numpy in Choregraphe for NAO robot?

43,351,742

0.197375

python,opencv,numpy,nao-robot,choregraphe

It depends if you're using a real NAO or a simulated one. Simulated one: choregraphe use its own embedded python interpreter, even if you add library to your system it won't change anything Real NAO: the system python interpreter is used, you need to install those library to your robot (and not to the computer running choregraphe). As pip ofthen doesn't work fine in NAO, you'll have to manually copy library to /home/nao/.local/lib/python2.7/site-packages

I'm doing a project that require cv2 and numpy in one of the scripts using choregraphe, but I get an error : No module named cv2/numpy. I think it is because choregraphe has its own python interpreter but I do not know how to install cv2 and numpy into the python of choregraphe. How can I do it?

0

1

1,416

0

43,388,797

0

2

false

0

2017-04-12T05:18:00.000

0

3

0

export geometric stiffness matrix in abaqus software

43,360,854

0

python,abaqus

Abaqus doesn't output stiffness matrices. If you already know Abaqus and only want to do buckling, you might want to try CalculiX. I think it can output a stiffness matrix and uses Abaqus-style input files.

I need export geometric(stress or diffrential) stiffness matrix in linear buckling problem at abaqus software for use in matlab program, i find method export of usual stiffness matrix, but i cant find any data about export geometric stiffness matrix in abaqus script, Can some one help me about it? for example when i export stiffness matrix in buckling problem,the result is same as stiffness matrix in linear static problem

0

1

797

0

43,375,546

0

2

false

0

2017-04-12T05:18:00.000

1

3

0

export geometric stiffness matrix in abaqus software

43,360,854

0.066568

python,abaqus

Let's discuss a little back ground on geometric nonlinear FEA. There are many different methods used for nonlinear geometric analysis of structure. The most well-known and established ones are [1]: Total Lagrangian (TL), Updated Lagrangian (UL). TL: Uses the full nonlinear definition of strains. UL: Uses the linear definition of strain but updates the reference configuration with previous deformation solutions. Abaqus's core solver uses a very specific type of geometric nonlinear algorithm called co-rotational in where the rotation and deformation with respect to reference configuration are treated separately. Abaqus's co-rotational formulation is propitiatory and I do not expect them to offer its output so easily. Even if you could see the output for nonlinear geometry matrices from Abaqus. These matrice(s) would be different than what you expect depending on the methods that you have used for nonlinear geometric modeling. [1] Reddy, Junuthula Narasimha. An Introduction to Nonlinear Finite Element Analysis: with applications to heat transfer, fluid mechanics, and solid mechanics. OUP Oxford, 2014.

I need export geometric(stress or diffrential) stiffness matrix in linear buckling problem at abaqus software for use in matlab program, i find method export of usual stiffness matrix, but i cant find any data about export geometric stiffness matrix in abaqus script, Can some one help me about it? for example when i export stiffness matrix in buckling problem,the result is same as stiffness matrix in linear static problem

0

1

797

0

43,383,011

0

1

false

5

2017-04-13T03:04:00.000

2

3

0

How to calculate the distance in meaning of two words in Python

43,382,857

0.132549

python,nlp,nltk

My suggestion is as follows: Put each word through the same thesaurus, to get a list of synonyms. Get the size of the set of similar synonyms for the two words. That is a measure of similarity between the words. If you would like to do a more thorough analysis: Also get the antonyms for each of the two words. Get the size of the intersection of the sets of antonyms for the two words. If you would like to go further!... Put each word through the same thesaurus, to get a list of synonyms. Use the top n (=5, or whatever) words from the query result to initiate a new query. Repeat this to a depth you feel is adequate. Make a collection of synonyms from the repeated synonym queries. Get the size of the set of similar synonyms for the two words from the two collections of synonyms. That is a measure of similarity between the words.

I am wondering if it's possible to calculate the distance/similarity between two related words in Python (like "fraud" and "steal"). These two words are not synonymous per se but they are clearly related. Are there any concepts/algorithms in NLP that can show this relationship numerically? Maybe via NLTK? I'm not looking for the Levenshtein distance as that relates to the individual characters that make up a word. I'm looking for how the meaning relates. Would appreciate any help provided.

0

1

2,056

0

43,403,593

0

1

true

1

2017-04-14T00:17:00.000

1

0

Reshape array in keras

43,403,560

1.2

python,keras,reshape

Just put the last dim to -1.. So Reshape(15,-1)

Is it possible in keras to [Reshape][1] an array in eras by only specifying one of the two dimension, such that the last dimension fits accordingly? In my case i have (30,1,2080) and i want to reshape it to (15,)

0

1

599

0

43,416,019

0

1

true

1

2017-04-14T12:42:00.000

3

1

0

tf.image.pad_to_bounding_box VS tf.pad and tf.image.crop_to_bounding_box VS tf.slice

43,411,738

1.2

python,tensorflow

Mostly you use them tf.image.* for easiness of use. Both crop_to_bounding_box and pad_to_bounding_box use slice and padunderneath, but also add checkings and constraints to make sure you don't spend hours trying to debug your slice/pad indices and offsets.

I'd like to understand why does the two functions tf.image.crop_to_bounding_box and tf.image.pad_to_bounding_box exists, since the behaviour of these two functions can be done really simply with respectively tf.slice and tf.pad. They are not so much easier to understand, and their scope is narrow since they accept only 3D and 4D tensors. Furthermore, they tend to be slower in terms of time of execution. Is there something I miss here ?

0

1

825

0

43,420,032

0

1

0

1

false

10

2017-04-14T13:59:00.000

6

2

0

Performance of pyomo to generate a model with a huge number of constraints

43,413,067

1

python,mathematical-optimization,pyomo

While you can use NumPy data when creating Pyomo constraints, you cannot currently create blocks of constraints in a single NumPy-style command with Pyomo. Fow what it's worth, I don't believe that you can in languages like AMPL or GAMS, either. While Pyomo may eventually support users defining constraints using matrix and vector operations, it is not likely that that interface would avoid generating the individual constraints, as the solver interfaces (e.g., NL, LP, MPS files) are all "flat" representations that explicit represent individual constraints. This is because Pyomo needs to explicitly generate representations of the algebra (i.e., the expressions) to send out to the solvers. In contrast, NumPy only has to calculate the result: it gets its efficiency by creating the data in a C/C++ backend (i.e., not in Python), relying on low-level BLAS operations to compute the results efficiently, and only bringing the result back to Python. As far as performance and scalability goes, I have generated raw models with over 13e6 variables and 21e6 constraints. That said, Pyomo was designed for flexibility and extensibility over speed. Runtimes in Pyomo can be an order of magnitude slower than AMPL when using cPython (although that can shrink to within a factor of 4 or 5 using pypy). At least historically, AMPL has been faster than GAMS, so the gap between Pyomo and GAMS should be smaller.

I am interested in the performance of Pyomo to generate an OR model with a huge number of constraints and variables (about 10e6). I am currently using GAMS to launch the optimizations but I would like to use the different python features and therefore use Pyomo to generate the model. I made some tests and apparently when I write a model, the python methods used to define the constraints are called each time the constraint is instanciated. Before going further in my implementation, I would like to know if there exists a way to create directly a block of constraints based on numpy array data ? From my point of view, constructing constraints by block may be more efficient for large models. Do you think it is possible to obtain performance comparable to GAMS or other AML languages with pyomo or other python modelling library ? Thanks in advance for your help !

0

1

4,313

0

43,507,382

0

1

true

3

2017-04-14T15:33:00.000

0

1

0

Undersampling vs class_weight in ScikitLearn Random Forests

43,414,689

1.2

python,scikit-learn,random-forest

I think you can split majority class in +-10000 samples and train the same model using each sample plus the same points of minority class.

I am applying ScikitLearn's random forests on an extremely unbalanced dataset (ratio of 1:10 000). I can use the class_weigth='balanced' parameter. I have read it is equivalent to undersampling. However, this method seems to apply weights to samples and do not change the actual number of samples. Because each tree of the Random Forest is built on a randomly drawn subsample of the training set, I am afraid the minority class will not be representative enough (or not representated at all) in each subsample. Is this true? This would lead to very biased trees. Thus, my question is: does the class_weight="balanced" parameter allows to build reasonably unbiased Random Forest models on extremely unbalanced datasets, or should I find a way to undersample the majority class at each tree or when building the training set?

0

1

900

0

53,412,662

0

1

0

4

false

26

2017-04-14T22:01:00.000

-1

13

0

how to install tensorflow on anaconda python 3.6

43,419,795

-0.015383

python,tensorflow,installation,python-wheel

conda create -n tensorflow_gpuenv tensorflow-gpu Or type the command pip install c:.*.whl in command prompt (cmd).

I installed the new version python 3.6 with the anaconda package. However i am not able to install tensorflow. Always receive the error that tensorflow_gpu-1.0.0rc2-cp35-cp35m-win_amd64.whl is not a supported wheel on this platform. How can I install tensorflow on anaconda (python 3.6)?

0

1

191,454

0

53,520,347

0

1

0

4

false

26

2017-04-14T22:01:00.000

0

13

0

how to install tensorflow on anaconda python 3.6

43,419,795

0

python,tensorflow,installation,python-wheel

Uninstall Python 3.7 for Windows, and only install Python 3.6.0 then you will have no problem or receive the error message: import tensorflow as tf ModuleNotFoundError: No module named 'tensorflow'

I installed the new version python 3.6 with the anaconda package. However i am not able to install tensorflow. Always receive the error that tensorflow_gpu-1.0.0rc2-cp35-cp35m-win_amd64.whl is not a supported wheel on this platform. How can I install tensorflow on anaconda (python 3.6)?

0

1

191,454

0

59,495,885

0

1

0

4

false

26

2017-04-14T22:01:00.000

1

13

0

how to install tensorflow on anaconda python 3.6

43,419,795

0.015383

python,tensorflow,installation,python-wheel

Well, conda install tensorflow worked perfect for me!

I installed the new version python 3.6 with the anaconda package. However i am not able to install tensorflow. Always receive the error that tensorflow_gpu-1.0.0rc2-cp35-cp35m-win_amd64.whl is not a supported wheel on this platform. How can I install tensorflow on anaconda (python 3.6)?

0

1

191,454

0

45,116,048

0

1

0

4

false

26

2017-04-14T22:01:00.000

-1

13

0

how to install tensorflow on anaconda python 3.6

43,419,795

-0.015383

python,tensorflow,installation,python-wheel

For Windows 10 with Anaconda 4.4 Python 3.6: 1st step) conda create -n tensorflow python=3.6 2nd step) activate tensorflow 3rd step) pip3 install --ignore-installed --upgrade https://storage.googleapis.com/tensorflow/windows/cpu/tensorflow-1.2.1-cp36-cp36m-win_amd64.whl

I installed the new version python 3.6 with the anaconda package. However i am not able to install tensorflow. Always receive the error that tensorflow_gpu-1.0.0rc2-cp35-cp35m-win_amd64.whl is not a supported wheel on this platform. How can I install tensorflow on anaconda (python 3.6)?

0

1

191,454

0

43,423,664

0

1

false

0

2017-04-15T02:20:00.000

0

1

0

Hdf5 and spatial indexes

43,421,508

0

python,pandas,quadtree,r-tree

Yes, r-trees can be stored on disk easily. (It's much harder with KD-trees and quad-trees) That is why the index is block oriented - the block size is meant to be chosen to match hour drive. I don't use pandas, and will not give a library recommendation.

I have a large dataset, 11 million rows, and I loaded the data into pandas. I want to then build a spatial index, like rtree or quad tree, but when I push it into memory, it consumes a ton of RAM along with the already reading the large file. To help reduce the memory footprint, I was thinking of trying to push the index to disk. Can you store the tree in a table? Or even a dataframe and store it in hdf table? Is there a better strategy? Thanks

0

1

245

0

43,425,655

0

1

0

1

true

2

2017-04-15T11:43:00.000

2

1

Is it recommended to use TensorFlow under Ubuntu or under Windows?

43,425,621

1.2

python,ubuntu,tensorflow,deep-learning

I think it's easier for you to use Ubuntu if you have the possibility. Getting lapack and blas libraries from sources is easier in linux (you can get precompiled packages for windows though). I prefer native pip, but for windows and for starting Anaconda should be the choice.

I am a newbie to TensorFlow (and the whole deep learning as well). I have a machine with dual boot, Windows 10 and Ubuntu 16. Under which OS should I install and run TensorFlow? Windows or Ubuntu? Also, what is the recommended Python environment? Anaconda or native pip?

0

1

8,589

0

43,436,325

0

1

false

0

2017-04-15T15:15:00.000

0

1

0

Curve fitting a sum of Gaussian's to 6 peaks

43,427,707

0

python,curve-fitting,gaussian

I have successfully used a genetic algorithm to search error space and find initial parameter estimates for scipy's Levenberg-Marquardt nonlinear solver. The most recent versions of scipy now include the Differential Evolution genetic algorithm which is what I had been using. It will take a bit of experimenting on a previously fit data set to tune things like population size or genetic crossover strategy, but then you should have a way for the computer to find initial parameter estimates in this way. It takes a while to run, so let it go overnight, but it can be automated.

I was wondering if anyone could show me a way of fitting multiple gaussian curves to a data set containing 6 peaks(data comes from a diffraction pattern from a copper gold alloy crystal). the way i have at the moment involves using multiple gaussian equations added together meaning i have to give multiple guesses of values when curve fitting.

0

1

523

0

43,432,668

0

1

0

1

true

2

2017-04-16T00:09:00.000

1

2

0

Chaotic permutations

43,432,439

1.2

python,algorithm

Use shuffle. It is fast enough and adequately chaotic for any practical purpose.

Is there any alorithm, which provides a chaotic permutation of a given list in a reasonable time? Im using Python, but I am concerned, that the given shuffle function will not provide a good solution, due a length of 1.1 Millions elements in the given list. I did some googling and did not find any usefull results, would be really surprised if there is anything like this, but I would really appreciate an answer

0

1

522

0

43,438,949

0

1

false

1

2017-04-16T15:10:00.000

2

0

Convert image file to float array in Python

43,438,653

0.197375

python

You can convert the list of the integers to list of floats as [float(i) for i in values] with list comprehension. An other option is to convert the img variable as numpy.ndarray to an other numpy.ndarray which contains float values: img = img.astype(float) After this assignment the results will contain float values.

How can I convert an image to an array of float numbers? img = cv2.imread('img.png') and now convert img to float so I get for print(img[0,0]) something like "[ 4.0 2.0 0.0] instead of [4 2 0] Do you have an idea? Thank you very much!

0

1

27,136

0

72,350,771

0

1

0

2

false

34

2017-04-16T18:48:00.000

1

3

0

The real difference between float32 and float64

43,440,821

0.066568

python,numpy,types

float32 is less accurate but faster than float64, and flaot64 is more accurate than float32 but consumes more memory. If speed accuracy is more important, you can use float64, and if speed is more important than accuracy, you can use float32.

I want to understand the actual difference between float16 and float32 in terms of the result precision. For instance, Numpy allows you to choose the range of the datatype you want (np.float16, np.float32, np.float64). My concern is that if I decide to go with float 16 to reserve memory and avoid possible overflow, would that create a loss of the final results comparing with float32 for instance? Thank you

0

1

53,683

0

52,804,163

0

1

0

2

false

34

2017-04-16T18:48:00.000

37

3

0

The real difference between float32 and float64

43,440,821

1

python,numpy,types

float32 is a 32 bit number - float64 uses 64 bits. That means that float64’s take up twice as much memory - and doing operations on them may be a lot slower in some machine architectures. However, float64’s can represent numbers much more accurately than 32 bit floats. They also allow much larger numbers to be stored. For your Python-Numpy project I'm sure you know the input variables and their nature. To make a decision we as programmers need to ask ourselves What kind of precision does my output need? Is speed not an issue at all? what precision is needed in parts per million? A naive example would be if I store weather data of my city as [12.3, 14.5, 11.1, 9.9, 12.2, 8.2] Next day Predicted Output could be of 11.5 or 11.5164374 do your think storing float 32 or float 64 would be necessary?

I want to understand the actual difference between float16 and float32 in terms of the result precision. For instance, Numpy allows you to choose the range of the datatype you want (np.float16, np.float32, np.float64). My concern is that if I decide to go with float 16 to reserve memory and avoid possible overflow, would that create a loss of the final results comparing with float32 for instance? Thank you

0

1

53,683

0

43,813,452

0

1

false

0

2017-04-16T20:30:00.000

0

1

0

How can I generate output predictions in Tensorflow just like model.predict in Keras?

43,441,827

0

python,numpy,tensorflow,ipython,keras

You can use tf.contrib.Keras to get same behavior.

Keras has model.predict which generates output predictions for the input samples. I am looking for this in tensorflow but cannot seem to find it or code it up.

0

1

288

0

43,459,094

0

1

false

0

2017-04-17T20:24:00.000

0

2

0

imread octave to imread cv2

43,459,032

0

python,numpy,image-processing,octave

What you have is a 3D array in octave. Here in the x-dimension you seem to have RGB values for each pixel and Y and Z dimension are the rows and columns respectively. However when you print it you will see all the values in the array and hence it looks like a 1D array.

I have to translate a code from Octave to Python, among many things the program does something like this: load_image = imread('image.bmp') which as you can see its a bitmap, then if I do size(load_image) that prints (1200,1600,3) which its ok, but, when I do: load_image it prints a one dimensional array, that does not make any sense to me, my question is how in Octave are these values interpreted because I have to load the same image in opencv and I couldn't find the way. thanks.

0

1

201

0

43,467,467

0

1

0

1

false

6

2017-04-18T07:54:00.000

-2

4

0

how to combine 2D arrays into a 3D array in python?

43,466,644

-0.099668

python,arrays

If your arrays are only 'list', sumplt defines an empty list at the beginning and append item into it: foo=[] for i in range(14): ... foo.append(tab)

I have a project in which there is a for loop running about 14 times. In every iteration, a 2D array is created with this shape (4,3). I would like to concatenate those 2D arrays into one 3D array (with the shape of 4,3,14) so that every 2D array would be in different "layer". How should that be implemented in Python?

0

1

8,835

0

43,482,677

0

1

0

1

true

0

2017-04-18T13:25:00.000

0

1

0

Python memory use for DataFrames in python script and in Anaconda Spyder

43,473,556

1.2

python,memory-management,anaconda,spyder

What is the effect in a script called something.py and executed as python something.py? Memory is unloaded after completion of execution. Please confirm. Yes, I think memory is freed after execution. What is the effect when I run something.py in say Anaconda Spyder. The spyder memory will not be unloaded unless I disconnect from the kernel. Is that a true statement? Spyder doesn't hold memory, only the IPython kernel associated with a console does. To free that memory, you need to restart the kernel.

Python 101 question. A pandas dataframe is created as: df1=pandas.DataFrame(data, index=index, columns=columns) # takes up say 100 MB memory now df2=df1 # will memory usage be doubled? What is the effect in a script called something.py and executed as python something.py? Memory is unloaded after completion of execution. Please confirm. What is the effect when I run something.py in say Anaconda Spyder. The spyder memory will not be unloaded unless I disconnect from the kernel. Is that a true statement? Thank you all for being patient with a Python newbie

0

1

844

0

43,481,217

0

1

0

1

false

1

2017-04-18T17:43:00.000

1

0

How to manage complexity while using IPython notebooks?

43,478,908

0.197375

python,ipython-notebook,code-organization,project-organization

Well, I have this problem now and then when working with a big set of data. Complexity is something I learned to live with, sometimes it's hard to keep things simple. What i think that help's me a lot is putting all in a GIT repository, if you manage it well and make frequent commits with well written messages you can track the transformation to your data easily. Every time I make some test, I create a new branch and do my work on it. If it gets to nowhere I just go back to my master branch and keep working from there, but the work I done is still available for reference if I need it. If it leads to something useful I just merge it to my master branch and keep working on new tests, making new branches, as needed. I don't think it answer all of your question and also don't know if you already use some sort version control in your notebooks, but that is something that helps me a lot and I really recommend it when using jupyter-notebooks.

Imagine that you working with a large dataset, distributed over a bunch of CSV files. You open an IPython notebook and explore stuff, do some transformations, reorder and clean up data. Then you start doing some experiments with the data, create some more notebooks and in the end find yourself heaped up with a bunch of different notebooks which have data transformation pipelines buried in them. How to organize data exploration/transformation/learning-from-it process in such a way, that: complexity doesn't blow, raising gradually; keep your codebase managable and navigable; be able to reproduce and adjust data transformation pipelines?

0

1

161

0

48,334,045

0

1

false

1

2017-04-18T23:14:00.000

0

2

0

Combining effects of dummy variables in a regression model

43,483,717

0

python,sas,regression,random-forest,dummy-variable

You may google SAS STAT manual /User guide. Check out any major regression procedures there that support Class statement. Underneath the Class it details Reference... option. They all detail how a design matrix is fashioned out. The way you fed your 100 dummies must have been obvious enough to trigger JMP to roll back to a temp class variable that re-engineer back to one single variable. If you want to know how exactly JMP is triggered to do roll back, go to JMP website and open a technical support track. But mechanically I am confident this is what happens.

I am building a regression model with about 300 features using Python Sklearn. One of the features has over 100 categories and I end up having ~100 dummy columns for this feature.Now each of the dummy column has its own coefficient, or a feature ranking score (if using Random Forest or xgb) - which is something I dont like. However, when I create the same model in SAS JMP, it gives me one single feature score for the feature with 100 categories -it apparently handles categories automatically. Can someone tell me how SAS JMP combines the coefficients/feature importances of 100 dummy variables into one metric. And how can i can achieve the same in Python.

0

1

284

0

43,486,149

0

1

false

0

2017-04-19T03:45:00.000

2

1

0

How to show only years on x axis of pandas histogram?

43,485,972

0.379949

python,pandas

There are two ways to do it. 1) if you just want to change the display format, by using the {:,.0f} format to explicitly display (rounded) floating point values with no decimal numbers: pd.options.display.float_format = '{:,.0f}'.format 2) if you want to convert it to int : df.col = df.col.astype(int)

I have a pandas histogram that shows the frequency that specific years show up in a dataframe. The x axis contains 2006.0, 2006.5, 2007.0, 2007.5, etc. However I want my histogram x axis to only have 2006, 2007, etc. This will make my histogram clearer, especially since in my df I only have values for particular years in their integer form, not 2006.5, 2007.4, etc. How would I go about doing this?

0

1

220

0

57,392,720

0

1

0

2

false

9

2017-04-19T16:19:00.000

2

7

0

anaconda cannot import matplotlib.pyplot

43,501,102

0.057081

python

Just open anaconda prompt and use either of the below command to install the package. This solved my issue. conda install -c plotly chart-studio or conda install -c plotly/label/test chart-studio

I am getting this error when I am trying to import "matplotlib.pyplot". I cant even install matplotlib.pyplot through conda install. It shows this: import matplotlib.pyplot Traceback (most recent call last): File "", line 1, in ModuleNotFoundError: No module named 'matplotlib.pyplot'

0

1

25,113

0

70,727,396

0

1

0

2

false

9

2017-04-19T16:19:00.000

0

7

0

anaconda cannot import matplotlib.pyplot

43,501,102

0

python

Check if the ...../python3.x/site-packages is listed within sys.path. If not append it with sys.path.append('.....python3.8/site-packages')

I am getting this error when I am trying to import "matplotlib.pyplot". I cant even install matplotlib.pyplot through conda install. It shows this: import matplotlib.pyplot Traceback (most recent call last): File "", line 1, in ModuleNotFoundError: No module named 'matplotlib.pyplot'

0

1

25,113

0

62,819,510

0

1

false

1

2017-04-20T08:04:00.000

1

0

ValueError: Clustering algorithm could not initialize. Consider assigning the initial clusters manually

43,513,662

0.197375

algorithm,python-3.x

Your data might not warrant a larger number of clusters. Run the algorithm for lesser number of k values and note the total cost at the end. If this stops decreasing, there is no need to increase k. It's called the elbow method, you can look it up.

I have data containing a mixture of numeric values and categorical values. I used K-prototype to cluster them. init = 'Huang' n_clusters = 50 max_iter = 100 kproto = kprototypes.KPrototypes(n_clusters=n_clusters,init=init,n_init=5,verbose=verbose) clusters = kproto.fit_predict(data_cats_matrix,categorical=categoricals_indicies) when I run the last code I'm getting an error as follows : ValueError: Clustering algorithm could not initialize. Consider assigning the initial clusters manually.

0

1

2,181

0

43,531,728

0

1

0

1

true

1

2017-04-20T23:40:00.000

1

0

Integer overflow in universal hashing implementation

43,531,607

1.2

python,numpy,hash

(x + y) % z == ((x % z) + (y % z)) % z. So you could take the modulus before doing the sum: Cast a and x to uint64. (Multiply two uint32 will never overflow uint64). Compute h = (a * x) % p + b Return (h - p) if h > p else h. (Alternatively: return h % p)

For one of my pet projects I would like to create an arbitrary number of different integer hashes. After a bit of research I figured that universal hashing is the way to go. However, I struggle with the (numpy) implementation. Say I'm trying to use the hash family h(a,b)(x) = ((a * x + b) mod p) mod m and my x can be anywhere in the range of uint32. Then choosing p >= max(x) and a, b < p means that in the worst case a * x + b overflows not only uint32 but also uint64. I tried to find an implementation which solves this problem but only found clever ways to speed up the modulo operation and nothing about overflow. Any suggestions for how to implement this are greatly appreciated. Thanks :-)

0

1

189

0

43,532,000

0

1

true

0

2017-04-21T00:21:00.000

2

1

0

Python Machine Learning classifier for just one label

43,531,951

1.2

python,machine-learning

The lack of negative examples does not make this unary classification; there is no such modelling, as one-class has no discrimination, and therefore derives no new information from the data set. As you've pointed out, there are two classes: Legitimate and not. That's binary. Use any binary classifier from your research that's capable of deriving boundaries from positive data only. For instance, so-called "one-class" SVM is one such classifier.

I'm new to machine learning and looking to do run a training / testing dataset through a few classifiers, but the problem I'm having is that I only have one label for my data (Legitimate, currently set as an int so 1 for legit, 0 for not). Ideally I'm looking for a classifier that is going to run with just one label and either confirm or deny if something falls into that label, without the need to specify a second label or class. Any help would be greatly appreciated! Many thanks.

0

1

438

0

43,544,802

0

1

false

1

2017-04-21T13:03:00.000

2

1

0

Dashboard Filter Superset

43,543,469

0.379949

python,superset

I found a possibility to create a fast filter. A table 'fcountry' is created, where all countries (DE,FR, etc.) are stored. This table is used to create the filter widget, which is added to the Dashboard. However, I'm still looking for a handy solution for EU without DE (where Country!='DE'). At the Moment I need to select all the countries except DE.

I have a test data table (called "eu") with around 5 millions Observations, where each Observation belongs a country in the EU. Now I'd like to implement a country filter on my dashboard for further Analysis. There already exists a filter widget, which does a query on my data set "eu" to get distinct Countries. However, I already know the Country names for the filter and I could skip (if possible) the query. Is there a possibility to set a filter without a query? The filter would look like: DE (where Country='DE') FR (where Country='FR') etc. EU without DE (where Country!='DE') Regards

0

1

1,424

0

43,576,690

0

1

0

5

false

9

2017-04-22T10:03:00.000

0

9

0

Jupyter can't find keras' module

43,557,881

0

python,anaconda,keras,jupyter-notebook

I have realized that I had two different Jupyter's directories, so I have manually deleted one on them. Finally, I have reinstalled Anaconda. Now Keras works properly.

I have installed Tensorflow and Keras by Anaconda (on Windows 10), I have created an environment where I am using Python 3.5.2 (the original one in Anaconda was Python 3.6). When I try to execute import keras as ks, I get ModuleNotFoundError: No module named 'keras'. I have tried to solve this issue by sys.path.append(C:\\Users\\ ... \\Anaconda3\\python.exe) with both notebook and console, but I continue to get the same error. How could I solve this issue?

0

1

27,759

0

43,557,975

0

1

0

5

false

9

2017-04-22T10:03:00.000

1

9

0

Jupyter can't find keras' module

43,557,881

0.022219

python,anaconda,keras,jupyter-notebook

(Not an answer but some troubleshooting hints) sys.path is not the path to your Python executable, but the path to the libraries. Check where Keras is installed and check your sys.path. How exactly did you install Keras? Try to execute the same command from the Python interpreter. Do you have the same issue? How did you install Jupiter, is the sys.path visible from there the same as sys.path visible from your Python interpreter? Do Jupiter and Keras use the same version of Python? You might try to uninstall Jupiter and install it again, and hope that the new installation picks up the packages which are already installed. What could happen is that you have more than one Python installation and different libraries being installed to the different places. sys.path, when requested from different environments, might give you a hint if that's true.

I have installed Tensorflow and Keras by Anaconda (on Windows 10), I have created an environment where I am using Python 3.5.2 (the original one in Anaconda was Python 3.6). When I try to execute import keras as ks, I get ModuleNotFoundError: No module named 'keras'. I have tried to solve this issue by sys.path.append(C:\\Users\\ ... \\Anaconda3\\python.exe) with both notebook and console, but I continue to get the same error. How could I solve this issue?

0

1

27,759

0

60,808,850

0

1

0

5

false

9

2017-04-22T10:03:00.000

0

9

0

Jupyter can't find keras' module

43,557,881

0

python,anaconda,keras,jupyter-notebook

Here's how I solved this problem. First, the diagnosis. When I run which python in a terminal window on my Mac (the same terminal I used to launch jupyter, I get /Users/myusername/.conda/envs/myenvname/bin/python, but when I run the same command from a terminal within Jupyter, I get /usr/bin/python. So Jupyter isn't using the correct python executable; the version it's using doesn't have any of my packages installed. But which jupyter returns /usr/bin/jupyter; it's using a version of jupyter that isn't coming from within my conda environment. I ran conda install jupyter and now which jupyter returns /Users/myusername/.conda/envs/myenvname/bin/jupyter (for some reason I had to restart the terminal window for this to take effect.) Then if I relaunch jupyter notebook, the notebook is using the correct version of Python and I have access to all my installed conda packages.

I have installed Tensorflow and Keras by Anaconda (on Windows 10), I have created an environment where I am using Python 3.5.2 (the original one in Anaconda was Python 3.6). When I try to execute import keras as ks, I get ModuleNotFoundError: No module named 'keras'. I have tried to solve this issue by sys.path.append(C:\\Users\\ ... \\Anaconda3\\python.exe) with both notebook and console, but I continue to get the same error. How could I solve this issue?

0

1

27,759

0

51,338,496

0

1

0

5

false

9

2017-04-22T10:03:00.000

0

9

0

Jupyter can't find keras' module

43,557,881

0

python,anaconda,keras,jupyter-notebook

If you are a windows/mac user who are working on Jupyter notebook “pip install keras” doesn't help you .Try the below steps.It was solved for me 1. In command prompt navigate to the “site packages” directory of your anaconda installed. 2. Now use “conda install tensorflow” and after “conda install keras” 3. Re-start your Jupyter notebook and run the packages.

I have installed Tensorflow and Keras by Anaconda (on Windows 10), I have created an environment where I am using Python 3.5.2 (the original one in Anaconda was Python 3.6). When I try to execute import keras as ks, I get ModuleNotFoundError: No module named 'keras'. I have tried to solve this issue by sys.path.append(C:\\Users\\ ... \\Anaconda3\\python.exe) with both notebook and console, but I continue to get the same error. How could I solve this issue?

0

1

27,759

0

56,765,580

0

1

0

5

false

9

2017-04-22T10:03:00.000

0

9

0

Jupyter can't find keras' module

43,557,881

0

python,anaconda,keras,jupyter-notebook

Acually, I did this command pip install keras and sudo -H pip3 install keras and pip3 install keras. None of them worked. I added the following command and everything worked like a charm: pip install Keras. Yes a capital 'K'

I have installed Tensorflow and Keras by Anaconda (on Windows 10), I have created an environment where I am using Python 3.5.2 (the original one in Anaconda was Python 3.6). When I try to execute import keras as ks, I get ModuleNotFoundError: No module named 'keras'. I have tried to solve this issue by sys.path.append(C:\\Users\\ ... \\Anaconda3\\python.exe) with both notebook and console, but I continue to get the same error. How could I solve this issue?

0

1

27,759

0

43,563,552

0

1

0

1

false

1

2017-04-22T19:05:00.000

1

0

How to iterate/ loop through a large (>2GB) JSON dataset in R/ Python?

43,563,447

0.197375

python,json,r,bigdata

The jsonlite R package supports streaming your data. In that way there is no need to read all the json data into memory. See the documentation of jsonlite for more details, the stream_in function in particular. Alternatively: I would dump the json into a mongo database and process the data from that. You need to install mongodb, and start running mongod. After that you can use mongoimport to import the json file into the database. After that, you can use the mongolite package to read data from the database.

I was trying to do some exploratory analyses on a large (2.7 GB) JSON dataset using R, however, the file doesn't even load in the first place. When looking for solutions, I saw that I could process the data in smaller chunks, namely by iterating through the larger file or by down-sampling it. But I'm not really sure how to do that with a JSON dataset. I also thought of converting the original JSON data into .csv, but after having a look around that option didn't look that helpful. Any ideas here?

0

1

609

0

69,907,112

0

1

false

7

2017-04-24T01:41:00.000

0

2

0

AttributeError: module 'numpy' has no attribute 'core'

43,578,533

0

python,numpy,pyspark,anaconda

Apart from upgrading and re-installing, sometimes it is caused by your Pandas. It might have dependency on older numpy so you may have to upgrade or reinstall pandas if upgrading numpy alone didn't resolve your problem.

I was wondering if anyone had this issue when running spark and trying to import numpy. Numpy imports properly in a standard notebook, but when I try importing it via a notebook running spark, I get this error. I have the most recent version of numpy and am running the most recent anaconda python 3.6. Thanks! --------------------------------------------------------------------------- AttributeError Traceback (most recent call last) in () ----> 1 import numpy /Users/michaelthomas/anaconda/lib/python3.6/site-packages/numpy/__init__.py in () 144 return loader(*packages, **options) 145 --> 146 from . import add_newdocs 147 __all__ = ['add_newdocs', 148 'ModuleDeprecationWarning', /Users/michaelthomas/anaconda/lib/python3.6/site-packages/numpy/add_newdocs.py in () 11 from __future__ import division, absolute_import, print_function 12 ---> 13 from numpy.lib import add_newdoc 14 15 ############################################################################### /Users/michaelthomas/anaconda/lib/python3.6/site-packages/numpy/lib/__init__.py in () 6 from numpy.version import version as __version__ 7 ----> 8 from .type_check import * 9 from .index_tricks import * 10 from .function_base import * /Users/michaelthomas/anaconda/lib/python3.6/site-packages/numpy/lib/type_check.py in () 9 'common_type'] 10 ---> 11 import numpy.core.numeric as _nx 12 from numpy.core.numeric import asarray, asanyarray, array, isnan, \ 13 obj2sctype, zeros AttributeError: module 'numpy' has no attribute 'core'

0

1

22,785

0

43,579,860

0

1

false

5

2017-04-24T04:05:00.000

0

3

0

Pandas Plot With Positive Values One Color And Negative Values Another

43,579,626

0

python,pandas,matplotlib,plot

Make 2 separate dataframes by using boolean masking and the where keyword. The condition would be if >0 or not. Then plot both datframes one by one ,one top of the other, with different parameters for the color.

I have a pandas dataframe where I am plotting two columns out the 12, one as the x-axis and one as the y-axis. The x-axis is simply a time series and the y-axis are values are random integers between -5000 and 5000 roughly. Is there any way to make a scatter plot using only these 2 columns where the positive values of y are a certain color and the negative colors are another color? I have tried so many variations but can't get anything to go. I tried diverging color maps, colormeshs, using seaborn colormaps and booleans masks for neg/positive numbers. I am at my wits end.

0

1

7,728

0

43,592,085

0

1

true

13

2017-04-24T15:04:00.000

4

2

0

Trained Machine Learning model is too big

43,591,621

1.2

python,machine-learning,pickle,random-forest

In the best case (binary trees), you will have 3 * 200 * (2^30 - 1) = 644245094400 nodes or 434Gb assuming each one node would only cost 1 byte to store. I think that 140GB is a pretty decent size in comparision.

We have trained an Extra Tree model for some regression task. Our model consists of 3 extra trees, each having 200 trees of depth 30. On top of the 3 extra trees, we use a ridge regression. We trained our model for several hours and pickled the trained model (the entire class object), for later use. However, the size of saved trained model is too big, about 140 GB! Is there a way to reduce the size of the saved model? Are there any configuration in pickle that could be helpful, or any alternative for pickle?

0

1

9,137

0

43,597,467

0

1

false

0

2017-04-24T19:34:00.000

0

1

0

python: one kernel density plot which includes multiple columns in a single dataframe

43,596,345

0

python,kernel-density

I did find how to work around by transforming the dataframe's columns into one single column. df.stack()

I need to make a single gaussian kernel density plot of a dataframe with multiple columns which includes all columns of the dataframe. Does anyone know how to do this? So far I only found how to draw a gaussian kernel plot of a single column with seaborn. ax = sns.kdeplot(df['shop1']) However, neither ax = sns.kdeplot(df)norax = sns.kdeplot(df['shop1','shop2]) do not work. Otherwise is there a workaround where I could transform the dataframe with shape df.shape(544, 33) to (17952, 2), by appending each columns to eachother? The dataframe includes normalized prices for one product, whereas each column represents a different seller and the rows indicate date and time of the prices.

0

1

465

0

43,612,292

0

1

false

0

2017-04-25T13:05:00.000

2

0

Python: linguistic normalization

43,611,550

0.197375

python,nlp,nltk

This is a bit more experimental, but another possibility is to use word embeddings. The words great and good should have similar occurrence contexts, so their vectors should be similar, you cluster your words like that, and aggregate them into the same word/concept. Of course this will greatly depend on the corpus and methods you use to generate the embeddings.

I'm looking for a lemmatization module/lib that will transfer a sentence like: "this is great" to "this is good". I'm familiar with some of the tools available in nltk such as stemming and lemmatization, however it's not exactly what I'm looking for My goal is to minimize the variety of ways saying the same thing.

0

1

593

0

43,625,960

0

1

false

1

2017-04-26T03:07:00.000

0

1

0

How to extract COMPLAINT features from texts in order to classify complaints from non-complaints texts

43,624,308

0

python,nlp,classification,feature-extraction,sentiment-analysis

I think you will find that bag-of-words is not so naive. It's actually a perfectly valid way of representing your data to give it to an SVM. If that's not giving you enough accuracy you can always include bigrams, i.e. word pairs, in your feature vector instead of just unigrams.

I have a corpus of around 6000 texts with comments from social network (FB, twitter), news content from general and regional news and magazines, etc. I have gone through first 300 of these texts and tag each of these 300 texts' content as either customer complaint or non-complaint. Instead of naive way of bag of words, I am wondering how can I accurately extract the features of these complaints and non-complaints texts? My goal is to use SVM or other classification algorithm/ library such as Liblinear to most accurately classify the rest of these texts as either complaint or non-complaint with the current training set of 300 texts. Is this procedure similar to sentiment analysis? If not, where should I start?

0

1

319

0

44,355,165

0

1

0

1

true

2

2017-04-26T03:41:00.000

1

0

Is scipy.misc.imread safe/efficient to run from multiple threads?

43,624,618

1.2

python,multithreading,scipy,multiprocessing,python-imaging-library

scipy.misc.imread is safe from multiple threads, but each call locks the global interpreter, so performance won't benefit from multithreading. It works well from multiprocessing, no unexpected issues.

I have set up a producer/consumer model using python Queues. In one producer I'm reading images using scipy.misc.imread. Reading images in one thread is not fast enough, it takes ~0.2s per image to read. About 20MB/sec reading from an SSD. I tried adding another identical thread using python's threading module. However the time spent in scipy.misc.imread increased by approximately double, causing the 2 threads to read images approximately as fast as the 1 did. I'm sure my SSD can handle 40MB/sec throughput, even with random reads. A dd write test shows 800MB+/sec write speeds. I am left wondering if scipy.misc.imread runs as a critical region among threads? Would I expect multiprocessing to avoid the problem?

0

1

186

0

43,637,652

0

1

false

4

2017-04-26T14:47:00.000

0

3

0

Getting worse result using keras 2 than keras 1

43,637,515

0

python,keras,mse

Probably some default values that have changed from Keras 1.2. You should check the default values for you 1.2 code, and set the same value for your new code.

I ran the same code (with the same data) on CPU first using keras 1.2.0 and then keras 2.0.3 in both codes keras is with TensorFlow backend and also I used sklearn for model selection, plus pandas to read data. I was surprised when I got the MSE(Mean squared error) of 42 using keras 2.0.3 and 21 using keras 1.2.0. Can someone pls explain to me why this is happening? Why I am getting more error using keras 2? Thanks PS. This result is after editing the code to keras 2 standard. for example in Dense I change keras 1 code to keras 2 standard.

0

1

417

0

46,969,923

0

1

0

1

false

2

2017-04-26T17:50:00.000

-1

1

0

Numexpr detecting number of threads less than number of cores

43,641,247

-0.197375

python,arrays,cluster-computing,python-multithreading,numexpr

I´m not sure, how numexpr actually works internally, when detect_number_of_threads is called, but maybe it reads out the number of threads that is available to openmp and not the number of threads that were locally set.

I am using numexpr for simple array addition on a remote cluster. My computer has 8 cores and the remote cluster has 28 cores. Numexpr documentation says that "During initialization time Numexpr sets this number to the number of detected cores in the system" But the cluster gives this output. detect_number_of_cores() = 28 detect_number_of_threads()=8 ALthough when I try to set the number of threads manually to something else(set_num_threads=20) , array operation seems to run faster. But detect_number_of_threads() still gives 8 as output. Is this a bug?

0

1

5,204

0

48,454,088

0

1

false

1

2017-04-27T06:26:00.000

2

1

0

How to add proper nouns as vocab to Spacy models?

43,650,119

0.379949

python,nlp,named-entity-recognition,spacy

I'm not sure exactly what you want to do... but below are solutions for a few possibilities. You have a complete list of the medicines/drugs of interest... i. and you want a special rule for tokenisation of these strings: Would not recommend this approach but in principle you could add special cases to the Tokenizer. ii. and you simply want to find them: Then you should use a Matcher to find them efficiently after you've created the Doc. Then you can do what you want with the spans that you find. Make them pronouns if you want. You don't have a complete list of the medicines/drugs of interest... i. but you want to generally identify them: Then you'll need to train an NER model to recognize them. To generate the training data you could take a whole bunch of papers and use a Matcher looking for known medicines/drugs to grab a whole bunch of example sentences. Then you train... Later edit: To be clear, the Vocab only stores strings and assigns them a unique ID for working with under the hood. It's the Tagger which decides whether it's a PRON or not.

I am using Spacy 1.8.0 with Python and I would like to use Spacy to do analysis on Medical Documents. There is a way off adding new entity types to spacy's named entity recognizer. However, is it possible to add the names of medicines/drugs as proper nouns to spcay's vocab? Or do they need to be added by training the spacy NER? Thanks

0

1

806

0

43,651,884

0

1

false

3

2017-04-27T07:54:00.000

3

2

0

How to Combine 2 integer columns in a dataframe and keep the type as integer itself in python

43,651,802

0.291313

python,pandas,dataframe

Maybe something like df[year] * 100 + df[month] would help.

I have 2 columns df[year] and df[month]. It has values ranging from 2000 to 2017 and month values 1 - 12. How to combine these to another column which would contain the combined output. Eg: Year Month Y0M 2000 1 200001 2000 2 200002 2000 3 200003 2000 10 200010 Note : there is a 0 added in between Year and Month in Y0M column, (only for single digit numbers and not double digit) Currently I am able to only do this by converting to string, but I want to retain them as type numbers

0

1

3,502

0

43,688,972

0

1

false

0

2017-04-27T14:45:00.000

0

1

0

Python py2exe executable silent crash with scipy.linalg or numpy.linalg

43,661,027

0

python,numpy,scipy,py2exe

I managed to solve this problem and wanted to post the solution. The link posted by J.J. Hakala, showed that an error with the same symptoms can occur with matplotlib because py2exe does not transfer some necessary dll's to the new dist directory. I was already transferring the dll's mentioned in that post manually, however by manually testing I discovered that mkl_avx.dll was also necessary for numpy.linalg. By simply copying this to the dist directory this problem was resolved.

I've been using py2exe to package some scripts as executable, which has worked well until this error. In one script I need to solve a straightforward system of linear equations. I've been doing this with scipy.linalg.lstsq. The problem is that any script I package with any scipy.linalg or numpy.linalg command crashes as soon as it gets to that line. The executable compiles with no errors. Everything runs as it should until it gets to a linalg command, then the executable just closes with no warning or error. No error log gets written. I've confirmed that it's always the linalg commands by manually writing log files with flags around the commands. I've used other scipy packages, primarily optimize, before with no issues. I've tried catching the error with a 'try' statement, but it crashes regardless. I can only think it's some kind of import error, but those normally throw errors when the executable is being built. The complete lack of any error or warning messages has me at a loss as to what could be causing this.

0

1

189

0

43,662,378

0

1

false

0

2017-04-27T15:12:00.000

0

1

0

Feeding a queue from a list of arrays of different lengths

43,661,651

0

python,input,tensorflow,queue,pipeline

I assume you're using an RNN here, for an RNN to take a variable length array you're going to need to pad them with zeros to the batch length and pass in a value that tells the RNN how long each sequence is. I'd suggest taking the same approach here. Pad them before passing them into the TF queue and pass in another value that indicates the unpadded length. If you're not using an RNN (what are you doing with variable length inputs then??) you must be processing them one-by-one (you didn't give us much detail there) so just break them up and send them into the queue one by one, read them off in TF as you need and recombine them with TF slicing and joining operations.

I have a list of arrays (could be numpy arrays or just list of int) of different lengths and I would like to feed it into some TensorFlow queue for testing a larger input pipeline -- instead of serializing them and re-read from file system. Is it possible?

0

1

119

0

43,733,509

0

2

false

0

2017-04-27T21:35:00.000

0

4

0

Fast algorithm to find the n1 th smallest number to n2 th smallest number in an array

43,668,307

0

python,algorithm,sorting,columnsorting

If n2 (and therefore n1) are both small, then you can find the n2 smallest elements and ignore the first n1 ones. These approaches are described by Arun Kumar and user448810 and will be efficient as long as n1 remains small. However, you may be describing a situation in which n1 (and therefore n2) can grow (perhaps even linearly with the overall list length) and it is only their difference n2-n1 which remains small. In this case you need a selection algorithm such as quickselect which will remain O(N) in this case.

I have an array, I attempt to find the n1 th smallest number until the n2 th smallest number in the array and store them in another array of size n2-n1+1. Here n2>n1 and both are realtively small compared to the size of the array (for example, the array size is 10000, n1=5, n2=20). I can sort this array first, and then retrieve the n1 th number, n1+1 th number until the n2 th number from the sorted array. But since n1 and n2 are usually relatively small compared to the size of the array, it is not necessary to sort the array completely. The algorithm should be able to stop in middle once reaches n2 I think. I wonder if there is any algorithm, maybe a modified version of certian sorting algorithm that is specifacally good (by good I mean fast) at this problem. You can either use Python code or pseudo code as an illustration, thanks!

0

1

115

0

43,670,278

0

2

false

0

2017-04-27T21:35:00.000

1

4

0

Fast algorithm to find the n1 th smallest number to n2 th smallest number in an array

43,668,307

0.049958

python,algorithm,sorting,columnsorting

Since, the N1 and N2 are really small compared to size of array letus say N. We can have an implementation in O(N2 * LogN) using min heap data structures. Steps Construct a min heap. Complexity of this operation will be O(N) Loop for a range of 0 to N2: Get the root element and call heapify. Ignore first N1 elements and return rest of the elements. Complexity of this step is O(1)+O(logN) Hence, overall we have O(N2 * logN)

I have an array, I attempt to find the n1 th smallest number until the n2 th smallest number in the array and store them in another array of size n2-n1+1. Here n2>n1 and both are realtively small compared to the size of the array (for example, the array size is 10000, n1=5, n2=20). I can sort this array first, and then retrieve the n1 th number, n1+1 th number until the n2 th number from the sorted array. But since n1 and n2 are usually relatively small compared to the size of the array, it is not necessary to sort the array completely. The algorithm should be able to stop in middle once reaches n2 I think. I wonder if there is any algorithm, maybe a modified version of certian sorting algorithm that is specifacally good (by good I mean fast) at this problem. You can either use Python code or pseudo code as an illustration, thanks!

0

1

115

0

43,670,482

0

1

false

0

2017-04-28T01:07:00.000

0

1

0

How to run python code at prescribed time and store output in database

43,670,334

0

python,database-design,web-scraping

In windows, you can use Task Scheduler or in Linux crontab. You can configure these to run python with your script at set intervals of time. This way you don't have a python script continuously running preventing some hangup in a single call from impacting all subsequent attempts to scrape or store in database. To store the data there are many options which could either be a flat file(write to a text file), save as a python binary(use shelf or pickle), or install an actual database like MySQL or PostgreSQL and many more. Google for info. Additionally an ORM like SQLAlchemy may make controlling, querying, and updating your database a little easier since it will handle all tables as objects and create the SQL itself so you don't need to code all queries as strings.

I have written a piece of python code that scrapes the odds of horse races from a bookmaker's site. I wish to now: Run the code at prescribed increasingly frequent times as the race draws closer. Store the scraped data in a database fit for extraction and statistical analysis in R. Apologies if the question is poorly phrased/explained - I'm entirely self taught and so have no formal training in computer science. I don't know how to tell a piece of python code to run itself every say n-minutes and I also have no idea how to correctly approach building such a data base or what factors I should be considering. Can someone point me in the right direction for getting started on the above?

0

1

73

0

44,968,742

0

1

false

0

2017-04-28T03:36:00.000

1

2

0

How to build a machine learning system (smart system) that updates automatically?

43,671,556

0.099668

python,machine-learning,artificial-intelligence,data-analysis

you are asking "Incremental way of learning " !! it's depends on what task are you doing. you can not apply incremental learning everywhere. the real problem is when we train , in the process of training , there are many other things that are related(inter dependent). which gets changed during a new cycle. so if you can make your algorithm independent , then you can do "incremental way of learning" !!

I am kinda a newbie in machine learning. I have done lots of learning related to data analysis and ML algorithms, and I've got very good results and understanding of the algorithms. However, my approaches are normally getting datasets, writing scripts or notebooks to solve the problems and that's pretty much it. Which means, no matter how fancy I create a script, I have to run the script everytime for each update, this is not enough. I think that machine learning is more than that, and there are already smart systems that are constantly updating their systems automatically, and I am really interested in that, but can't really find any good materials that dive deep into that field. Let's just say, like the classic example, there's a website that has a ML system at the backend that identifies iris categories, and it's visitors constantly upload new observations of irises and the visitors get the category of the iris they upload, and the backend system continues to improve accordingly. How can I get to learn thing like this? Would be super keen to dive into this field. I'd really appreciate all your opinions! Thanks a lot!

0

1

405

0

44,010,620

0

1

false

1

2017-04-28T03:55:00.000

1

0

Extracting weights and tree structures from xgboost - plot trees

43,671,704

0.197375

python,algorithm,tree,random-forest,xgboost

You can dump the xgboost model into a text file, and parse it yourself. The file looks like this: booster[0(<-tree id)]: 0(<-node id):[f317(<-feature name)<0.187154] yes=1(<-child node id),no=2,missing=1 1:[f317<0.071172] yes=3,no=4,missing=3 ... 6379:leaf=0.125 (<-weight to that leaf) At the end, it is the weighted sum of all leaves. This is in case of binary classification and regression.

I'm using xgboost package in python and I'm going to extract tree structures after training. For instance, I'm going to know the feature and threshold that has been selected at each node to export the tree structures to a function. Moreover, I need to know the weight of each tree after training. (As we know, the result of trees will be combined like w1*h1+w2h2+...+wn*hn, in which wi is weigh and hi is the answer of ith tree in boosting). Actually, I need to have access to weights and inner trees. I'm doing classification. My other question is that, when I use "from xgboost import plot_tree", I receive this error "the package does not have this function". How can I plot my trees? Thanks

0

1

1,889

0

43,674,438

0

1

false

0

2017-04-28T07:24:00.000

0

2

0

I need to implement a SparseArray with a linked list implementation

43,674,393

0

python,arrays,algorithm,linked-list

I'd bet these are already implemented in numpy or scipy.

sparse array objects will have a fixed size n set when they are created - attempting to set or get elements larger than the size of the array should raise an IndexError

0

1

238

0

54,512,055

0

1

false

5

2017-05-01T18:45:00.000

0

3

0

Rendering a 3D mesh into an image using python

43,724,600

0

python,3d,rendering,render

nVidia's iRay is excellent however, 50k plus triangles is a serious task for any rendering engine. If you're looking for photorealistic quality you'll be looking at several minutes per render at 50k triangles from a simple lighting source. The more complex your lighting becomes the slower the process. A simple texture based shadow render is much quicker and can be done in real time, the results are limited to the quality of the UV mapped texture. Do you capture the surface material from the single image or do you apply a pre-mapped UV surface material to your standard 50k triangle model mesh?

I'm a working a neural network that does face reconstruction from a single image using tensorflow. I'm trying to figure out how to render the output of the network (which is a 3D mesh) into an image in python. Most of the libraries I found do 3D rendering in real time, I'm only looking to render and single frame. I also need something that is fast, because this is potentially going to be part of the pre-processing for the network in real-time (apply the network to live video. The mesh is quite simple, it's a simple face with about 30k vertices and 50k triangles.

0

1

11,128

0

43,735,408

0

1

true

1

2017-05-02T05:33:00.000

1

0

How to avoid padded indices during calculation of softmax using tf.nn.softmax?

43,730,740

1.2

python-3.x,tensorflow,deep-learning

You could set your padded cells to a value close to -Infinity (i.e. very small compared to your other logits), so that their contributions are just negligible.

I have a matrix where each row is padded to maxlen. If I apply softmax on the matrix, the padded indices are also considered. How to apply softmax on the matrix while considering only the non-padded indices? Note: padded length in each row of the matrix would differ.

0

1

302

0

65,209,790

0

1

false

1

2017-05-02T13:51:00.000

0

1

opencv_annotation not recognised by windows

43,739,684

0

windows,python-2.7,opencv

You need to build it using CMake.

I am using OpenCV in windows. When I am using this command in windows command prompt then I am getting an error as opencv_annotation is not recognised as the internal or external command. opencv_annotation --annotations=/C:\Users\harsh\Desktop/annotation.txt --images=C:\Users\harsh\Desktop/pos/

0

1

743

0

44,855,960

0

1

false

0

2017-05-03T20:06:00.000

0

1

0

ValueError: unsupported pickle protocol: 3, python 2.7 cannot load pickle file even dump with protocol = 2

43,769,071

0

python,pickle

It is more likely to be a pickle file created with python 3. I guess you mainly use python 2.7. Either you re-create the pickle file with python 2.7 or you use python 3.

I have used: pickle.dump(data, f, protocol=2) And try to open the pickle file with python 2.7, however, it still pop up with the error "ValueError: unsupported pickle protocol: 3".

0

1

2,769

0

57,201,952

0

1

false

2

2017-05-03T20:24:00.000

0

1

0

Selecting test and train sets in Keras for Python

43,769,357

0

python,deep-learning,classification,keras,conv-neural-network

Your method is not impossible, but it is impractical. A Neural Network learns from the training dataset. If you give the same data again in the test set, it will generate accurate predictions, which will have no relation with the network's predictive power.

I am trying to use Keras in Python for time series classification. I have made a simple CNN and was wondering if it is possible to randomly select a subsection of the full data set at each epoch to serve as the test set? I am quite new to neural nets so if there are any reasons this is impossible or impractical I would appreciate clarity.

0

1

95

0

43,772,528

0

1

false

106

2017-05-04T01:22:00.000

7

5

0

How to print a specific row of a pandas DataFrame?

43,772,362

1

python,python-3.x,pandas,indexing

Sounds like you're calling df.plot(). That error indicates that you're trying to plot a frame that has no numeric data. The data types shouldn't affect what you print(). Use print(df.iloc[159220])

I have a massive dataframe, and I'm getting the error: TypeError: ("Empty 'DataFrame': no numeric data to plot", 'occurred at index 159220') I've already dropped nulls, and checked dtypes for the DataFrame so I have no guess as to why it's failing on that row. How do I print out just that row (at index 159220) of the data frame? Thanks

0

1

543,751

0

43,775,529

0

1

0

1

false

1

2017-05-04T06:09:00.000

2

3

0

Store points by x, y and multiple z

43,775,108

0.132549

python,arrays,list,dictionary,store

Here are a couple of ways of doing this: As Geotop suggested, use a dictionary indexed with a tuple. This according to me is an elegant solution. Depending on the purpose and ease of traverse, you may convert the same to a nested list as follows: a = [ [ (1,2), (3,4), (5,6) ], [ [123,234,345], [567,678,543], [433,645,754] ] ] Now you may access the list of z values corresponding to (x,y) using a[1][i], where the tuple (x,y) is stored as a[0][i]. The second method seems much better to me as it avoids dictionary use and makes indexing much easier (as it is indexed with integers.). But at the end, it all depends on your purpose. Storing it randomly in a list and getting the values by a search is a bad idea if you look up the data frequently or if the data is huge.

I will write an example so you can understand me : For instance, I have these 3 points : (0, 1, 244) - (0, 1, 255) - (1, 2, 133) Actually I need to average the Z when 2 points have the same (x, y). My idea is to store them in something (an array, a dictionary ?) with a double index/key where the value is an array of the Z. So I basically want to have this : [0, 1, [244, 255]] and [1, 2, [133]] My problem is I don't know what to use to store them the simpliest way to do so... Thanks in advance !

0

1

85

0

68,590,093

0

1

false

14

2017-05-04T08:06:00.000

2

3

0

How to split a DataFrame in pandas in predefined percentages?

43,777,243

0.132549

python-3.x,pandas

Creating a dataframe with 70% values of original dataframe part_1 = df.sample(frac = 0.7) Creating dataframe with rest of the 30% values part_2 = df.drop(part_1.index)

I have a pandas dataframe sorted by a number of columns. Now I'd like to split the dataframe in predefined percentages, so as to extract and name a few segments. For example, I want to take the first 20% of rows to create the first segment, then the next 30% for the second segment and leave the remaining 50% to the third segment. How would I achieve that?

0

1

22,451

0

43,782,019

0

2

false

1

2017-05-04T11:28:00.000

0

2

0

Python 3.5 32 & 64 bit -- Scipy & Tensorflow issues

43,781,633

0

python,python-3.x,tensorflow,scipy,64-bit

Did you try using Anaconda or similar for the installation? - if this is an option in your case I would highly recommend it under Windows.

I'm trying to install Python 3.5 both 32 & 64 bit and also be able to transfer between the two as needed, but am not having any luck. Scipy will only install when I use the 32bit (Various issues when trying to install 64bit version even with physical .whl files). Meanwhile Tensorflow only works on x64. I'm using windows 7, and have tried various solutions I've found on Stackoverflow, but have had no luck. Also, was thinking of just dual installing linux mint and running python off there. Would any of you recommend? Thank you!

0

1

384

0

43,782,112

0

2

false

1

2017-05-04T11:28:00.000

0

2

0

Python 3.5 32 & 64 bit -- Scipy & Tensorflow issues

43,781,633

0

python,python-3.x,tensorflow,scipy,64-bit

Installing Linux (highly recommended) While your Issue with Scipy & Tensorflow. Why don't you use the Anaconda Installer for Windows x64 bit(it contains alomost all major packges for Data Science already(Scipy too)) , and then install Tensorflow

I'm trying to install Python 3.5 both 32 & 64 bit and also be able to transfer between the two as needed, but am not having any luck. Scipy will only install when I use the 32bit (Various issues when trying to install 64bit version even with physical .whl files). Meanwhile Tensorflow only works on x64. I'm using windows 7, and have tried various solutions I've found on Stackoverflow, but have had no luck. Also, was thinking of just dual installing linux mint and running python off there. Would any of you recommend? Thank you!

0

1

384

0

46,379,569

0

1

false

6

2017-05-05T04:53:00.000

1

2

1

Loading local file from client onto dask distributed cluster

43,796,774

0.099668

python,python-3.x,dask

Network solution : Under Windows only it should works with a shared forlder: dd.read_csv("\\server\shared_dir") Under Unix/Linux only it should works with HDFS: import hdfs3 and then hdfs.read_csv('/server/data_dir'...) But if you want to use Windows AND Linux workers at the same time I don't know since dd.read_csv() with UNC does not seem to be supported under Linux (because of the file path '\server\data_dir') and HDFS with hdfs.read_csv is not working under Windows (import hdfs3 failed because the lib libhdfs3.so doesn't exist under Windows) Does anyone have a Network solution for workers under Windows and Unix ?

A bit of a beginner question, but I was not able to find a relevant answer on this.. Essentially my data about (7gb) is located on my local machine. I have distributed cluster running on the local network. How can I get this file onto the cluster? The usual dd.read_csv() or read_parquet() fails, as the workers aren't able to locate the file in their own environments. Would I need to manually transfer the file to each node in the cluster? Note: Due to admin restrictions I am limited to SFTP...

0

1

2,868

0

43,798,752

0

1

false

24

2017-05-05T06:54:00.000

0

4

0

One-Hot-Encode categorical variables and scale continuous ones simultaneouely

43,798,377

0

python,scikit-learn

Can't get your point as OneHotEncoder is used for nominal data, and StandardScaler is used for numeric data. So you shouldn't use them together for your data.

I'm confused because it's going to be a problem if you first do OneHotEncoder and then StandardScaler because the scaler will also scale the columns previously transformed by OneHotEncoder. Is there a way to perform encoding and scaling at the same time and then concatenate the results together?

0

1

26,114

0

43,805,354

0

2

true

1

2017-05-05T11:51:00.000

1

2

0

Tensor-flow training and Testing on different OS

43,804,250

1.2

linux,windows,python-3.x,tensorflow

No, the model will be exactly the same. You'll only have to make sure that your TF versions on Linux and Windows are compatible ones, but this is not made more difficult by the different OS, it's only a matter of which versoin you install on which device.

Will the performance of the model be affected if I train the data on LINUX system and then use that model in the WINDOWS application or a python script?

0

1

496

0

43,805,554

0

2

false

1

2017-05-05T11:51:00.000

0

2

0

Tensor-flow training and Testing on different OS

43,804,250

0

linux,windows,python-3.x,tensorflow

Akshay, as gdelab said, make sure about the version. If you are running the same model on same data(training and testing) again & again, it might affect but using on different OS might not. I faced the same issue.

Will the performance of the model be affected if I train the data on LINUX system and then use that model in the WINDOWS application or a python script?

0

1

496

0

43,895,291

0

1

true

0

2017-05-06T13:43:00.000

0

1

0

How does one input images and labels for Semantic Instance Segmentation with neural networks?

43,821,167

1.2

python,machine-learning,tensorflow,conv-neural-network,image-segmentation

You'll want to train your NN in such as way that you'll be able to use it for prediction. If you want to just predict the classes from the image, then all you want to send to your NN is the original image (probably color balanced) and predict the classes from the XML (convert that into a 1 hot class encoding) if you want to predict the mask also, then send the original image (probably color balanced) and predict the mask and classes from the XML (convert that into a 1 hot class encoding) The above objectives (just classes, or classes+mask prediction) drives the decision to store the classes or the classes + mask.

So I know for a standard convolutional neural network you can provide the neural net (NN) a file with a list of labels or simply separate your classes by folders but for instance segmentation I imagine it's different right? For example using a site like labelme2 you can annotate and segment images and then download them along with mask files and XML files for labels. Does one need to subsequently input the original image, mask image and XML file to the instance segmentation NN? Thanks in advance.

0

1

667

0

43,835,158

0

1

false

1

2017-05-07T18:15:00.000

0

3

0

How to change data types "object" in Pandas dataframe after importing a CSV?

43,835,016

0

python-3.x,pandas

During reading a csv file: Use dtype or converters attribute in read_csv in pandas import pandas as pd import numpy as np df = pd.read_csv('data.csv',dtypes = {'a':float64,'b':int32},headers=None) Here,automatically the types will be read as the datatype you specified. After having read the csv file: Use astype function to change the column types. Check this code. Consider you have two columns df[['a', 'b']] = df[['a', 'b']].astype(float) The advantage of this is you change type of multiple columns at once.

I have imported a CSV file as a Pandas dataframe. When I run df.dtypes I get most columns as "object", which is useless for taking into Bokeh for charts. I need to change a column as int, another column as date, and the rest as strings. I see the data types only once I import it. Would you recommend changing it during import (how?), or after import?

0

1

7,317

0

68,993,454

0

1

false

1

2017-05-08T01:33:00.000

0

1

0

Draw the plot result in an image and save it as an image in python

43,838,497

0

python,image,plot

Use the command fig.savefig('line plot.jpg', bbox_inches='tight', dpi=150)

I want to draw the plot result in an image and save it as an image in python. I have a method that could Harris method in result there are alot of point that I want to trace in a picture but win I save the figure is giving me just the point not the image also please give me an advice

0

1

40

0

43,879,111

0

1

false

21

2017-05-09T00:18:00.000

3

2

0

Keras - is it possible to view the weights and biases of models in Tensorboard

43,859,133

0.291313

python,tensorflow,keras,tensorboard

I debugged this and found that the problem was I was not providing any validation data when I called fit(). The TensorBoard callback will only report on the weights when validation data is provided. That seems a bit restrictive, but I at least have something that works.

I just got started with Keras and built a Q-learning example program. I created a tensorboard callback and I include it in the call to model.fit, but the only things that appear in TensorBoard are the scalar summary for the loss and the network graph. Interestingly, if I open up the dense layer in the graph, I see a little summary icon labeled "bias_0" and one labeled "kernel_0", but I don't see these appearing in the distributions or histograms tabs in TensorBoard like I did when I built a model in pure tensorflow. Do I need to do something else to enable these in Tensorboard? Do I need to look into the details of the model that Keras produces and add my own tensor_summary() calls?

0

1

17,275

0

43,860,724

0

1

true

0

2017-05-09T02:17:00.000

0

1

0

Python - convert mysql timestamps type to matplotlib and graph

43,859,988

1.2

python,mysql,matplotlib

you can use datetime module,although i use now() function to extract datetime from mysql,but i consider the format is the same。 for instance python>import datetime as dt i put the datetime data into a list named datelist,and now you can use datetime.strptime function to convert the date format to what you want python>dates = [dt.datetime.strptime(d,'%Y-%m-%d %H:%M:%S') for d in datelist At last,you can put the list named dates into plot X-axis I hope it helps you.

After doing a bit of research I am finding it difficult to find out how to use mysql timestamps in matplotlib. Mysql fields to plot X-axis: Field: entered Type: timestamp Null: NO Default: CURRENT TIMESTAMP Sample: 2017-05-08 18:25:10 Y-axis: Field: value Type: float(12,6) Null: NO Sample: 123.332 What date format is matplotlib looking for? How do I convert to this format? I found out how to convert from unix timestamp to a format that is acceptable with matplotlib, is unix timestamp better than the timestamp field type I am using? Should I convert my whole table to unix timestamps instead? Would appreciate any help!

0

1

202

0

43,870,685

0

2

false

2

2017-05-09T12:21:00.000

1

3

0

dynamically growing array in numba jitted functions

43,869,734

0.066568

python,numpy,dynamic-arrays,numba

Typically the strategy I employ is to just allocate more than enough array storage to accommodate the calculation and then keep track of the final index/indices used, and then slice the array down to the actual size before returning. This assumes that you know beforehand what the maximum size you could possibly grow the array to is. The thought is that in most of my own applications, memory is cheap but resizing and switching between python and jitted functions a lot is expensive.

It seems that numpy.resize is not supported in numba. What is the best way to use dynamically growing arrays with numba.jit in nopython mode? So far the best I could do is define and resize the arrays outside the jitted function, is there a better (and neater) option?

0

1

2,018