Fairness via In-Processing in the Over-parameterized Regime: A Cautionary Tale with MinDiff Loss

Publication
Transactions of Machine Learning Research