search

glm-tools / pyglmnet

Python implementation of elastic-net regularized generalized linear models
http://glm-tools.github.io/pyglmnet/
MIT License
279 stars 83 forks source link

GLM parameter random_state ignored #269

Closed geektoni closed 5 years ago

geektoni commented 6 years ago

I noticed that with GLM I cannot get consistent and replicable results even if I set the random_state parameter. Different run of the same code will produce always different results.

I think that the problem can be found at this line: https://github.com/glm-tools/pyglmnet/blob/6763a9550fe87f1ab099db4f575a9e449e0b6a2d/pyglmnet/pyglmnet.py#L678 Since the RandomState object is never used to initialize randomly some of the parameters of the model, the random_state becomes useless.

jasmainak commented 6 years ago

indeed @geektoni do you want to help fix it with a pull request?

geektoni commented 6 years ago

Sure! #270 contains a proposed fix.

jasmainak commented 5 years ago

closed by #270