Byzantine-Robust Decentralized Learning via ClippedGossip

Author name not available (Why is that?)

Publication date: 3 February 2022

Abstract: In this paper, we study the challenging task of Byzantine-robust decentralized training on arbitrary communication graphs. Unlike federated learning where workers communicate through a server, workers in the decentralized environment can only talk to their neighbors, making it harder to reach consensus and benefit from collaborative training. To address these issues, we propose a ClippedGossip algorithm for Byzantine-robust consensus and optimization, which is the first to provably converge to a

O (d e l t a_{m a x} z e t a^{2} / g a m m a^{2})

neighborhood of the stationary point for non-convex objectives under standard assumptions. Finally, we demonstrate the encouraging empirical performance of ClippedGossip under a large number of attacks.

Has companion code repository: https://github.com/epfml/byzantine-robust-decentralized-optimizer

This page was built for publication: Byzantine-Robust Decentralized Learning via ClippedGossip

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6390026)