Analyzing the Regularizing Effects of Group-Fair Training on Shared Models

The author explores the benefits of group-fair training in reducing overfitting and improving generalization error for minority groups, providing refined bounds for individual group risks.