[MRG] MAINT sync/add support sample_weight with liblinear#15038
[MRG] MAINT sync/add support sample_weight with liblinear#15038NicolasHug merged 15 commits intoscikit-learn:masterfrom
Conversation
|
@amueller you might want to have a look at that |
adrinjalali
left a comment
There was a problem hiding this comment.
Not a complete review, but looks pretty good.
|
@adrinjalali any other comments? |
jnothman
left a comment
There was a problem hiding this comment.
I can confirm the tests fail in master...
adrinjalali
left a comment
There was a problem hiding this comment.
nits, otherwise, as far as I can tell, this looks good.
jnothman
left a comment
There was a problem hiding this comment.
I've checked the C code looks reasonably safe.... and that the tests are valid... I'd do better to compare to upstream, but I'm okay to see this merged.
|
Just realized this kinda needs a whats_new entry. |
Co-Authored-By: Nicolas Hug <contact@nicolas-hug.com>
|
Thanks @glemaitre ! |
|
Sweet! |
|
Is this functionality in any of the stable sklearn releases? What should I upgrade to? |
|
Not yet. The new release should be out this week |
|
Also, would anyone be keen on updating the old docs to mention the fact that |
Since this is a bug, it would be better to backport the fix. But we will probably release tomorrow, not sure that it would be worth. |
With the latest release of sklearn, we have merged scikit-learn#15038 which officially adds support for sample weights to these models. Let's update the docs to reflect this.
With the latest release of sklearn, we have merged scikit-learn#15038 which officially adds support for sample weights to these models. Let's update the docs to reflect this.
Add support for
sample_weightfor the loss/penalty in liblinear.Hopefully it will close:
closes #10873
closes #15018