Roman Lutz

I understand this as follows with this example screenshot:  The top right corner has a selection for "Color value". If I select "True Y" or "Predicted Y" I can't...

This reminds me somewhat of https://research.google.com/bigpicture/attacking-discrimination-in-ml/ I read this like @riedgar-ms and assumed that you want to threshold on the probabilities (or as you called them: "confidence scores"). You can...

That's a great question! I'm actually not entirely sure. @fairlearn/fairlearn-maintainers wdyt? Have you tried out whether it's technically possible to add more than one `versionchanged` tag? If it's not possible...

I agree. We've had some changes in the datasets module which @adrinjalali and I recommended omitting from `versionchanged` since only small changes or doc changes happened. Example: from the above...

No worries! You can always check your branch with `git status`, ideally before pushing 🙂 If you pushed to `main` on your fork you can always copy those changes to...

ExponentiatedGradient and GridSearch are done thanks to @bramreinders97! The other files under fairlearn.reductions remain, of course.

We may want to separate this into several issues (one per algorithm) if it turns out to be a lot of work. I do expect there to be a lot...

@SeanMcCarren it's definitely still relevant! I would suggest starting with one algorithm only and trying it out with LabelEncoder. Thanks for volunteering!

I agree with @riedgar-ms and suspect the former is much easier. I'm a little confused as to why this broke in the first place. What changed?

- using TF or PyTorch in tests is a MUCH easier task than in the docs because we'd have to include these packages in the dev requirements to make them...