Abstract

Non-IID data present a tough challenge for federated learning. In this paper,we explore a novel idea of facilitating pairwise collaborations between clientswith similar data. We propose FedAMP, a new method employing federatedattentive message passing to facilitate similar clients to collaborate more. Weestablish the convergence of FedAMP for both convex and non-convex models, andpropose a heuristic method to further improve the performance of FedAMP whenclients adopt deep neural networks as personalized models. Our extensiveexperiments on benchmark data sets demonstrate the superior performance of theproposed methods.

 

Quick Read (beta)