English

Robust Multi-Agent Bandits Over Undirected Graphs

Machine Learning 2023-01-30 v2 Multiagent Systems Machine Learning

Abstract

We consider a multi-agent multi-armed bandit setting in which nn honest agents collaborate over a network to minimize regret but mm malicious agents can disrupt learning arbitrarily. Assuming the network is the complete graph, existing algorithms incur O((m+K/n)log(T)/Δ)O( (m + K/n) \log (T) / \Delta ) regret in this setting, where KK is the number of arms and Δ\Delta is the arm gap. For mKm \ll K, this improves over the single-agent baseline regret of O(Klog(T)/Δ)O(K\log(T)/\Delta). In this work, we show the situation is murkier beyond the case of a complete graph. In particular, we prove that if the state-of-the-art algorithm is used on the undirected line graph, honest agents can suffer (nearly) linear regret until time is doubly exponential in KK and nn. In light of this negative result, we propose a new algorithm for which the ii-th agent has regret O((dmal(i)+K/n)log(T)/Δ)O( ( d_{\text{mal}}(i) + K/n) \log(T)/\Delta) on any connected and undirected graph, where dmal(i)d_{\text{mal}}(i) is the number of ii's neighbors who are malicious. Thus, we generalize existing regret bounds beyond the complete graph (where dmal(i)=md_{\text{mal}}(i) = m), and show the effect of malicious agents is entirely local (in the sense that only the dmal(i)d_{\text{mal}}(i) malicious agents directly connected to ii affect its long-term regret).

Keywords

Cite

@article{arxiv.2203.00076,
  title  = {Robust Multi-Agent Bandits Over Undirected Graphs},
  author = {Daniel Vial and Sanjay Shakkottai and R. Srikant},
  journal= {arXiv preprint arXiv:2203.00076},
  year   = {2023}
}
R2 v1 2026-06-24T09:57:01.061Z