Abstract: Personalized Federated Learning (PFL) is primarily de-signed to provide customized models for each client to better fit the non-iid distributed client data, which is a inherent challenge in ...