When this is fixed for feature_selection.from_model._LearntSelectorMixin, that should be modified to include mixin feature_selection.base.SelectorMixin.

Can I take up this issue ?

Hi, it seems that the issue has been fixed for AgglomerativeClustering. But the issue seems to be still persisting for _LearntSelectorMixin and seems for SparsePCA in here. In case anyone is working on this, I would gladly step aside !

In the _LearntSelectorMixin, the threshold parameter is right now passed to the transform function. When is mentioned that

should have selection_threshold parameter on object

is it expected to be a class variable for this Mixin or the base class SelectorMixin ?

Also there is a comment mentioning that

Note because of the extra threshold parameter in transform, this does
not naturally extend from SelectorMixin

Pardon my ignorance regarding the structure of classes, but which would be the best place to put the threshold option ?

Thanks.

_LearntSelectorMixin is deprecated, so not applicable either.

On 4 January 2016 at 01:06, Maniteja Nandana [email protected]
wrote:

Hi, it seems that the issue has been fixed for AgglomerativeClustering.
But the issue seems to be still persisting for _LearntSelectorMixin and
seems for SparsePCA in here
https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/decomposition/sparse_pca.py#L133.
In case anyone is working on this, I would gladly step aside !

In the _LearntSelectorMixin, the threshold parameter is right now passed
to the transform function. When is mentioned that

should have selection_threshold parameter on object

is it expected to be a class variable for this Mixin or the base class
SelectorMixin ?

Also there is a comment mentioning that

Note because of the extra threshold parameter in transform, this does not
naturally extend from SelectorMixin

Pardon my ignorance regarding the structure of classes, but which would be
the best place to put the threshold option ?

Thanks.

—
Reply to this email directly or view it on GitHub
#1975 (comment)
.

Thanks. I realized that it was not exposed on the public API but thought that it needs to be fixed since it is being used in other places. Sorry for not thinking over it.

@jnothman should we close this? Or first check if any models still have that with inspect?

Apart from a number of copy instances, a git grep 'def transform' quickly shows one remaining:

sklearn/decomposition/sparse_pca.py:    def transform(self, X, ridge_alpha=None):