DRFLM: Distributionally Robust Federated Learning with Inter-client Noise via Local Mixup

The authors propose a general framework to solve the two challenges simultaneously: inter-client data heterogeneity and intra-client data noise. The framework uses distributionally robust optimization to mitigate the negative effects caused by data heterogeneity and incorporates mixup techniques into the local training process to mitigate the effects of intra-client data noise.