[MRG] Pass sample_weight when predicting on stacked folds

[wderose]

Reference Issues/PRs

Fixes #16537.

What does this implement/fix? Explain your changes.

This PR passes sample_weight via fit_params when we call cross_val_predict in _BaseStacking, as per the comment here.

scikit-learn/sklearn/ensemble/_stacking.py

Lines 169 to 175 in 136ef79

	predictions = Parallel(n_jobs=self.n_jobs)(
	delayed(cross_val_predict)(clone(est), X, y, cv=deepcopy(cv),
	method=meth, n_jobs=self.n_jobs,
	verbose=self.verbose)
	for est, meth in zip(all_estimators, self.stack_method_)
	if est != 'drop'
	)

Any other comments?

cc: @caioaao @glemaitre

[jnothman]

Please add a |Fix| entry to the change log at doc/whats_new/v0.23.rst. Like the other entries there, please reference this pull request with :pr: and credit yourself (and other contributors if applicable) with :user:

Please add a versionchanged note in the docstring of sample_weight

[jnothman]

But tests are failing. Should have checked that first

[wderose]

But tests are failing. Should have checked that first

@jnothman : Thank you for the prompt review.

I've added the requested doc changes and fixed the failing specs.

Please let me know if you'd like more details in the versionchange/release notes or if you'd like to see more tests.

[glemaitre]

This looks good. Only a couple of comments

[wderose]

@glemaitre Thank you for the comments.

I updated the release notes to reference StackingClassifier and StackingRegressor directly.

I also created a follow-on issue (#16595) regarding the naming of _parallel_fit_estimator and added a comment noting the strange naming.

Please take another look at these changes and let me know if there is anything else you'd like me to include.

[wderose]

I needed to add this call so that n_features_in_ is set correctly in CheckingClassifier.

The property is read during the fit method of BaseStacking: d205638#diff-c8f397939f2393da5b2eef4285761da1R143. We use CheckingClassifier in the unit tests for the stacking estimators: https://github.com/scikit-learn/scikit-learn/pull/16539/files#diff-3979301645c7cbae79edc4085d997029R443

[thomasjpfan]

Thank you for the PR @wderose

LGTM!