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
48 stars 34 forks source link

SLEP022: Fixing randomness ambiguities #88

Open NicolasHug opened 1 year ago

NicolasHug commented 1 year ago

This SLEP is NOT the same as its predecessor https://github.com/scikit-learn/enhancement_proposals/pull/24

Disclaimer: I won't be able to champion this SLEP. I'm opening it here now because I hope it can help better framing the discussions happening in https://github.com/scikit-learn/scikit-learn/issues/26148.

betatim commented 1 year ago

I'd like to take on this SLEP if it is Ok with you.

NicolasHug commented 1 year ago

Sure, happy for you to take over @betatim , thanks!

joelostblom commented 1 year ago

Thanks for working on this! I think these clarification in the docs and the updated behavior would be really helpful to reduce ambiguity. I noticed that in https://github.com/scikit-learn/scikit-learn/issues/26148#issuecomment-1536979006 point 3 there was a mention that the new NumPy Generators would support a different behavior than the old RandomState objects (if I understood correctly). I didn't see a mention of that distinction in this proposal, is this difference still planned to be implemented?