-
-
Notifications
You must be signed in to change notification settings - Fork 26.2k
Description
In commit 6769490, the feature_selection module was restructured to make the constituent modules (univariate_selection, variance_threshold, rfe, from_model, mutual_info, and base) into part of the private API (by prepending them with underscores). The most important elements of feature_selection are, of course, exported through __init__.py and are part of the public API. However, in this restructuring, the SelectorMixin class was moved from the public to the private API. SelectorMixin was always unusual compared to other mixins, which typically live within the sklearn.base module (e.g., ClassifierMixin, TransformerMixin, RegressorMixin, etc.). But now that sklearn.feature_selection.base has been depreciated, SelectorMixin is the only mixin that's part of the private API.
My proposed solution would be to import SelectorMixin in feature_selection/__init__.py, as well as adding it to the __all__ list for that module.