You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Hi fairlearn community! My colleague and I recently did some work on multiclass fairness I thought might be worth sharing here. In our paper, we extended Hardt et al. 2016's notions of equal opportunity and equalized odds (along with a few other fairness measures) to the multiclass setting, and we did some small experiments to show how the method works under different data scenarios, like high class imbalance or high disparity across groups. We also have what I'd call a research-grade implementation of the method in Python on our GitHub page, if anyone's interested in seeing how the experimental results were generated. The method might be worth rolling into fairlearn at some point, since I've seen a handful of users asking about multiclass problems, but, as I've seen several contributors mention in response, notions of multiclass fairness can be unintuitive and/or tricky to apply, and so it might take some care in presenting, if it were included. :)
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
Hi fairlearn community! My colleague and I recently did some work on multiclass fairness I thought might be worth sharing here. In our paper, we extended Hardt et al. 2016's notions of equal opportunity and equalized odds (along with a few other fairness measures) to the multiclass setting, and we did some small experiments to show how the method works under different data scenarios, like high class imbalance or high disparity across groups. We also have what I'd call a research-grade implementation of the method in Python on our GitHub page, if anyone's interested in seeing how the experimental results were generated. The method might be worth rolling into fairlearn at some point, since I've seen a handful of users asking about multiclass problems, but, as I've seen several contributors mention in response, notions of multiclass fairness can be unintuitive and/or tricky to apply, and so it might take some care in presenting, if it were included. :)
Beta Was this translation helpful? Give feedback.
All reactions