Roman Lutz

I *think* this is because `DemographicParity` looks at difference in the selection rate between each group and the overall selection rate, so that can be below your threshold, say `0.05`,...

I marked it as a feature-request. I cannot provide an estimate at this point, but will update as soon as this gets scheduled.

@Pooja1301 is looking into this now for `ExponentiatedGradient`, and then same resampling should be applicable to `GridSearch` subsequently. Stay tuned :-)

@violazhong @riedgar-ms and I discussed this earlier today. Capturing some of our notes. The most important aspect of this work is passing in the resampling method. @riedgar-ms pointed out that...

All of this sounds reasonable and is fairly close to @violazhong 's current implementation. @violazhong please let us know if any of this doesn't make sense. Happy to elaborate!

This isn't a maintained project, so I expect that you'd need to make several changes to make it work with latest package versions. I didn't have the good sense to...