Accelerating Federated Edge Learning

Tuan Dung Nguyen, Amir Rezaei Balef, Canh T. Dinh, Nguyen Hoang Tran, Duy T. Ngo, Tuan Anh Le, Phuong Luu Vo

Abstract: Transferring large models in federated learning (FL) networks is often hindered by clients’ limited bandwidth. We propose $\textsf {FedAA}$ , an FL algorithm which achieves fast convergence by exploiting the regularized Anderson acceleration (AA) on the global level. First, we demonstrate that FL can benefit from acceleration methods in numerical analysis. Second, $\textsf {FedAA}$ improves the convergence rate for quadratic losses and improves the empirical performance for smooth and strongly convex objectives, compared to FedAvg, an FL algorithm using gradient descent (GD) local updates. Experimental results demonstrate that employing AA can significantly improve the performance of FedAvg, even when the objective is non-convex.