Leveraging Function Space Aggregation for Federated Learning at Scale

arXiv:2311.10291v2 Announce Type: replace
Abstract: The federated learning paradigm has motivated the development of methods for aggregating multiple client updates into a global server model, without sharing client data. Many federated learning algorithms, including the canonical Federated Averaging (FedAvg), take a direct (possibly weighted) average of the client parameter updates, motivated by results in distributed optimization. In this work, we adopt a function space perspective and propose a new algorithm, FedFish, that aggregates local approximations to the functions learned by …

abstract aggregation algorithms arxiv canonical client cs.lg data development federated learning function global multiple paradigm scale server space type updates