Multi-Agent Adversarial Training Using Diffusion Learning

Ying Cao; Elsa Rizk; Stefan Vlaski; Ali H. Sayed

ICASSP 2023accepted0 citations

Multi-Agent Adversarial Training Using Diffusion Learning

Ying Cao, Elsa Rizk, Stefan Vlaski, Ali H. Sayed

Abstract

This work focuses on adversarial learning over graphs. We propose a general adversarial training framework for multi-agent systems using diffusion learning. We analyze the convergence properties of the proposed scheme for convex optimization problems, and illustrate its enhanced robustness to adversarial attacks.

BibTeX

@inproceedings{icassp2023_multiagentadvers,
  title = {Multi-Agent Adversarial Training Using Diffusion Learning},
  author = {Ying Cao and Elsa Rizk and Stefan Vlaski and Ali H. Sayed},
  booktitle = {ICASSP 2023},
  year = {2023}
}