Multi-Agent Adversarial Training Using Diffusion Learning
arXiv:2303.01936v14 citationsh-index: 87
Originality Incremental advance
AI Analysis
This is an incremental improvement for multi-agent systems facing adversarial threats.
The paper tackles adversarial learning in multi-agent systems by proposing a diffusion learning framework, showing convergence for convex problems and improved robustness to attacks.
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.