10000 Add sample_weight support to binning in HGBT · Issue #27117 · scikit-learn/scikit-learn · GitHub
[go: up one dir, main page]

Skip to content
Add sample_weight support to binning in HGBT #27117
Open
@lorentzenchr

Description

@lorentzenchr

Use sample_weight in the binning of HistGradientBoostingClassifier and HistGradientBoostingRegressor, or allow it via an option.

Currently, sample weights are ignored in the _BinMapper.

Some more context and history summarized by @NicolasHug here:

I agree that it would make sense to support SW in the binner (although this is nuanced, see note below). Reading back the original PR, this was discussed extensively:

The estimators were still experimental at the time so we had more flexibility. Now that they're stable, bringing SW support in the Binner might require BC mitigations.

Also as a side note, #15657 (comment) may be relevant here: it's not super clear to me how we should handle SW in an estimator that performs some sort of subsambling during the training process (as is the case here during Binning).

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      0