Mention other, preferred, packages in docs

koaning / scikit-lego

Extra blocks for scikit-learn pipelines.

https://koaning.github.io/scikit-lego/

MIT License

1.24k stars 117 forks source link

Mention other, preferred, packages in docs #615

Open FBruzzesi opened 7 months ago

FBruzzesi commented 7 months ago

QuantileRegressor made it into scikit-learn some time ago already.

As in the future other features could lose the experimental status and be incorporated as well (I am looking at you TunedThresholdClassifier) should we plan to deprecate them in future versions of scikit-lego?

koaning commented 7 months ago

I've been wondering about this myself as well. The sklearn spline transformer now also does the repeating basis function trick ... so technically we won't need to have it around here.

Then again ... I also don't mind to brag. I think it was hosted on the lego side first, so maybe a message that just says "we're keeping it around for historical reasons because sklearn has support for it now" is fine? I don't mind to communicate that sklego is a testing ground for experimental ideas.

FBruzzesi commented 7 months ago

It seems fair and square! I will add a couple of info notes in the docstrings just to let users know.

koaning commented 7 months ago

I'll write a list of items that might fall under this category, feel free to add.

[ ] Some of our fairness docs should point to fairlearn
[ ] The quantile docs should point to sklearn
[ ] The repeating basis trick should point to the spline transformer in sklearn.