Refactor make_scorer for group fairness metrics

[chkoar]

No description provided.

[hoffmansc]

I like it!

Could you also adjust the docs? Just replace the two lines with make_scorer.

[chkoar]

Hey @hoffmansc, I think that addressed your comments. Is there something else?

Apart from that I think that the default scikit-learn scorers could be patched (probably using a patch function) so one could run:

cross_val_score(model, X, y, scoring="statistical_parity_difference")

or getting the scorer from sklearn.metrics.get_scorer just passing the string:

from sklearn.metrics import get_scorer
scorer = get_scorer("statistical_parity_difference")

Monkey-patching might be a controversial thing in the community.
If this is not the case, I could add this functionality in this or another PR.

[hoffmansc]

LGTM

[hoffmansc]

@chkoar I don't feel that strongly about it but I do think it might add a little bit of confusion about what functionality is coming from which library... For now, the current syntax doesn't seem too cumbersome to me

[chkoar]

but I do think it might add a little bit of confusion about what functionality is coming from which library

This might be true. In any case I opened an issue in the scikit-learn's upstream repo.