search

scikit-learn / enhancement_proposals

Enhancement proposals for scikit-learn: structured discussions and rational for large additions and modifications
https://scikit-learn-enhancement-proposals.readthedocs.io/en/latest
BSD 3-Clause "New" or "Revised" License
47 stars 34 forks source link

SLEP needed: slicling pipelines #13

Closed amueller closed 5 years ago

amueller commented 5 years ago

There's some implementations on slicing and subsetting pipelines but there's no agreement. We need a slep to discuss. https://github.com/scikit-learn/scikit-learn/pull/8431 https://github.com/scikit-learn/scikit-learn/pull/8448 https://github.com/scikit-learn/scikit-learn/issues/8414

raamana commented 5 years ago

Just out of curiosity, what would be the utility or applications for this feature be?

amueller commented 5 years ago

running the preprocessing pipeline so you know what the input for your model looks like. Or get the feature names for the features that go into the model. I added two related issues.

jnothman commented 5 years ago

Applying inverse_transform to the preprocessing pipeline without the last step is a common application.

But generally, it's something that's relatively easy to do with the current Pipeline implementation (Pipeline(my_pipeline.steps[:-1])), but with a proposed Pipeline implementation that clones steps when fitting, this would be harder

adrinjalali commented 5 years ago

Should we close this now that https://github.com/scikit-learn/scikit-learn/pull/2568 is merged?

amueller commented 5 years ago

yes.