Byzantine-Resilient Zero-Order Optimization for Communication-Efficient Heterogeneous Federated Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Egger, Maximilian, Bakshi, Mayank, Bitar, Rawad
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909471866880000
author Egger, Maximilian
Bakshi, Mayank
Bitar, Rawad
author_facet Egger, Maximilian
Bakshi, Mayank
Bitar, Rawad
contents We introduce CyBeR-0, a Byzantine-resilient federated zero-order optimization method that is robust under Byzantine attacks and provides significant savings in uplink and downlink communication costs. We introduce transformed robust aggregation to give convergence guarantees for general non-convex objectives under client data heterogeneity. Empirical evaluations for standard learning tasks and fine-tuning large language models show that CyBeR-0 exhibits stable performance with only a few scalars per-round communication cost and reduced memory requirements.
format Preprint
id arxiv_https___arxiv_org_abs_2502_00193
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Byzantine-Resilient Zero-Order Optimization for Communication-Efficient Heterogeneous Federated Learning
Egger, Maximilian
Bakshi, Mayank
Bitar, Rawad
Machine Learning
Cryptography and Security
Distributed, Parallel, and Cluster Computing
We introduce CyBeR-0, a Byzantine-resilient federated zero-order optimization method that is robust under Byzantine attacks and provides significant savings in uplink and downlink communication costs. We introduce transformed robust aggregation to give convergence guarantees for general non-convex objectives under client data heterogeneity. Empirical evaluations for standard learning tasks and fine-tuning large language models show that CyBeR-0 exhibits stable performance with only a few scalars per-round communication cost and reduced memory requirements.
title Byzantine-Resilient Zero-Order Optimization for Communication-Efficient Heterogeneous Federated Learning
topic Machine Learning
Cryptography and Security
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2502.00193