Adjusting test_size doesn't actually change test_size

[tyler-lanigan-hs]

Hello!
I'm trying to use this code for a project, however, I don't want my test size to be 0.5. When I try and adjust it, I don't get a change:

# from iterstrat.ml_stratifiers import MultilabelStratifiedShuffleSplit
import numpy as np

X = np.array([[1,2], [3,4], [1,2], [3,4], [1,2], [3,4], [1,2], [3,4]])
y = np.array([[0,0], [0,0], [0,1], [0,1], [1,1], [1,1], [1,0], [1,0]])
msss = MultilabelStratifiedShuffleSplit(n_splits=3, test_size=0.25, random_state=42)

for train_index, test_index in msss.split(X, y):
    print("TRAIN:", train_index, "TEST:", test_index)
    print(len(train_index))
    print(len(test_index))
    X_train, X_test = X[train_index], X[test_index]
    y_train, y_test = y[train_index], y[test_index]

outputs:

('TRAIN:', array([1, 2, 4, 7]), 'TEST:', array([0, 3, 5, 6]))
4
4
('TRAIN:', array([2, 3, 6, 7]), 'TEST:', array([0, 1, 4, 5]))
4
4
('TRAIN:', array([0, 2, 4, 6]), 'TEST:', array([1, 3, 5, 7]))
4
4

Koodos on putting this out there!

[trent-b]

Thank you for reporting an issue. However, I am unable to reproduce the issue. I believe I may have had some local changes that I had not committed to the repo, so I created a new version (0.1.6). I updated the files for pip and conda installation as well. Please try the updated version and let me know how it goes. Here is the output that I get:

TRAIN: [0 1 2 4 5 7] TEST: [3 6]
6
2
TRAIN: [0 1 3 4 5 7] TEST: [2 6]
6
2
TRAIN: [0 1 3 4 5 6] TEST: [2 7]
6
2

[tyler-lanigan-hs]

Thanks Trent. It's all good for me now :)

[mosheliv]

I still get the same behavior in 1.6

Collecting iterative-stratification
Collecting scipy (from iterative-stratification)
  Using cached https://files.pythonhosted.org/packages/45/d1/7c2b33a5daee3d67752d043fe7e1476c4465788b0b6e59367fd71fdf684a/scipy-1.2.0-cp27-cp27mu-manylinux1_x86_64.whl
Collecting numpy (from iterative-stratification)
  Using cached https://files.pythonhosted.org/packages/de/37/fe7db552f4507f379d81dcb78e58e05030a8941757b1f664517d581b5553/numpy-1.15.4-cp27-cp27mu-manylinux1_x86_64.whl
Collecting scikit-learn (from iterative-stratification)
  Using cached https://files.pythonhosted.org/packages/9e/29/bbf3414ba3d03cf1f8d8516e56d69e44ec0ad3fc79a3713b1c6809070e7d/scikit_learn-0.20.2-cp27-cp27mu-manylinux1_x86_64.whl
Installing collected packages: numpy, scipy, scikit-learn, iterative-stratification
Successfully installed iterative-stratification-0.1.6 numpy-1.14.5 scikit-learn-0.20.2 scipy-1.1.0
You are using pip version 8.1.1, however version 18.1 is available.
You should consider upgrading via the 'pip install --upgrade pip' command.
m@dl4:~/hpa$ python t.py 
('TRAIN:', array([1, 2, 4, 7]), 'TEST:', array([0, 3, 5, 6]))
4
4
('TRAIN:', array([2, 3, 6, 7]), 'TEST:', array([0, 1, 4, 5]))
4
4
('TRAIN:', array([0, 2, 4, 6]), 'TEST:', array([1, 3, 5, 7]))
4
4

[trent-b]

My intuition is that an earlier version of iterative-stratification is being used (even though I see that pip said that it installed 0.1.6). Try the following to confirm which version is being used:

import iterstrat
print(iterstrat.__version__)

Fix suggestions:

1. I suggest upgrading pip with pip install --upgrade pip and then installing.

2. If the upgrade does not resolve the issue, then try pip install iterative-stratification --no-cache-dir.

3. If the --no-cache-dir option does not work, then consider manually deleting the cache files as described here and then installing.

Please let me know if this works for you.

[mosheliv]

Found when it happens. Python 2.7 vs 3. Can you please give it a try in 2.7? not sure what needs to be changed for it to work in 2.7, probaby one of the default casts when dividing, this is the most common problem.

[trent-b]

The package doesn't currently support running under Python 2.7, only 3.4, 3.5, and 3.6 (see the Requirements section of README.md). I'll take a look though, but it may be a few weeks.

[mosheliv]

I managed to get around it by adding the future division and importing directly, so i am good for now! Thank you for this fantastic package

…

On Sun, Dec 23, 2018, 10:46 trent-b ***@***.*** wrote: Reopened #3 <#3> . — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#3 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AL6LC4sh5vEW_NllqZaDzBBX7PeGFIQSks5u7qgwgaJpZM4Vk0uD> .

[trent-b]

I'm glad you were able to find a workaround for 2.7.

[trent-b]

I've decided to stay with this package's claim that it only supports 3.X.