bramreinders97

Hi @romanlutz, after diving in to all the changes made in different versions, I am wondering what level of detail is expected in the versionchanged part of this issue. To...

Turns out it is possible to include multiple `versionchanged` tags. So I'll try to add all mentioned changes.

So I made some first changes and to test I tried committing to my local fork. I'm not sure if I accidentally pushed it to the main branch, as I...

For some reason my @fairlearn/fairlearn-maintainers tag did not work, so I'll do it like this: @adrinjalali @hildeweerts @MiroDudik @mmadaio @riedgar-ms @romanlutz

I would like to work on this issue :)

> @bramreinders97: could you perhaps write a short summary of the technique such that @fairlearn/fairlearn-maintainers can get an idea of how the method works? > Reject Option Classification is a...

In the paper, the authors propose the following API:  , where the big difference with `ThresholdOptimizer` is that the user is responsible for choosing which group(s) is/are in the...

> Thanks for the APIs! I'd imagine the theta would be set in the constructor or learned during `fit`, rather than provided in `predict`. Is there a `fit` method? Theta...

>The way I understood before was that we want to adjust close to the decision boundary (0.5), so for one group we lower it and for another we increase it,...

So taking the above into consideration, the following would be the API I'd suggest at this point: ```python fairlearn.postprocessing.RejectOptionClassifier( estimator , thresholds ) ``` With `estimator` = A sklearn estimator...