[SciPy-User] lstsq/Scipy and python multiprocessing

[Oleksandr Huziy]

Hi:

This code gives me the same answer in both cases.

[ -2.85439413e+14   2.11436602e+11]
[ -2.85439413e+14   2.11436602e+11]



sklearn.__version__ =  '0.14.1'
multiprocessing.__version__ = '0.70a1'

Cheers



2013/12/5 Charles Pierre <af.charles.pierre at gmail.com>

> I was trying to do some simple multivariate regression using
> sklearn.linear_model and mutliprocessing module when i found this really
> confusing behavior.
>
> For some reason, the linear regression seems to be broken for particular
> input vectors when using multiprocessing. Using the same training set
> without multiprocessing yields correct values ...
>
> Here a piece of code that demonstrates this weird behavior:
>
> import multiprocessingfrom sklearn import linear_model
> def test_without_multi(input_x,input_y):
>     clf = linear_model.LinearRegression(normalize=True)
>     clf.fit(input_x,
>             input_y,
>             n_jobs=1)
>     print clf.coef_
> def test_with_multi(input_x,input_y):
>     process = multiprocessing.Process(target=test_without_multi,args=(input_x,input_y))
>     process.start()
>     process.join()
> if __name__ == '__main__':
>     input_x = [[0,0],[0,0],[0,0],[0,0],[0,0],[0,0],[0,0],[0,0],[1,1350]]
>     input_y = [2,1,1,2,3,1,3,2,1]
>     test_without_multi(input_x,input_y)
>     test_with_multi(input_x,input_y)
>
> Does anyone know what is happening ?
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
>


-- 
Sasha
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.scipy.org/pipermail/scipy-user/attachments/20131205/8c3646b5/attachment.html>