Federated Learning Meets Multi-Objective Optimization

Federated learning has emerged as a promising, massively distributed way to train joint deep model over large amounts of edgedevices while keeping private user data strictly on device. In this work, motivated from ensuring fairness among users and robustness against malicious adversaries, we formulate federated multi-objective optimization propose new algorithm FedMGDA+ that is guaranteed converge Pareto stationary solutions. simple implement, fewer hyperparameters tune, refrains sacrificing the performance any participating user. We establish convergence properties point out its connections existing approaches. Extensive experiments variety datasets confirm compares favorably state-of-the-art.