FedMoS: Taming Client Drift in Federated Learning with Double Momentum and Adaptive Selection | IEEE Conference Publication | IEEE Xplore