Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Decentralized Multi-Agent Linear Bandits with Safety Constraints
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use triggered rare communication phases instead of communicating every round | DLUCB · Decentralized linear bandits | communication: \mathcal{O}\big(dN^2T\frac{\log(Nd)}{{\log^{0.5}(1/|\lambda_2|)}}\big) -> \mathcal{O}\big(d^3N^{2.5}\frac{\log(Nd)}{{\lo | Helped |
| Use triggered rare communication phases instead of communicating every round | DLUCB · Decentralized linear bandits | regret: \mathcal{O}\big(d\frac{\log(Nd)}{{\log^{0.5}(1/|\lambda_2|)}}\log(NT)+d\log(NT)\sqrt{NT}\big) -> \mathcal{O}\big(Nd^{1.5 | Hurt |
| Collaborate by sharing information across agents rather than learning independently | DLUCB · Synthetic linear bandits, Ring, Star, Complete, and random Erdős–Rényi graphs | regret | Helped |
| Use a graph with smaller spectral magnitude |λ₂| | DLUCB · Synthetic linear bandits, varying network topology | regret | Helped |
| Use RC-DLUCB rather than DLUCB | DLUCB · Synthetic linear bandits, varying network topology | regret | Hurt |
| Increase the number of agents | DLUCB · Synthetic random graphs with N=5, 10 and 15 | per-agent regret | Helped |
| Use the alternative consensus scheme that communicates aggregate Gram matrices and reward vectors (ablation) | DLUCB · Decentralized linear bandits | regret | No clear effect |
| Add safety constraints and use an enlarged reward confidence set | DLUCB · Safe decentralized linear bandits | regret: 2e\beta_{T}\sqrt{2dNT\log(\lambda+\frac{NT}{d})} -> 2e\kappa_r\beta_{T}\sqrt{2dNT\log(\lambda+\frac{NT}{d})} | Hurt |
Closest papers by meaning
- Decentralized Cooperative Stochastic Bandits (NeurIPS 2019 · similarity 0.95)
- Decentralized Randomly Distributed Multi-agent Multi-armed Bandit with Heterogeneous Rewards (NeurIPS 2023 · similarity 0.93)
- Distributed Multi-Agent Bandits Over Erdős-Rényi Random Networks (NeurIPS 2025 · similarity 0.92)
- Cooperative Multi-Agent Bandits with Heavy Tails (ICML 2020 · similarity 0.92)
- Distributed Contextual Linear Bandits with Minimax Optimal Communication Cost (ICML 2023 · similarity 0.91)
- Individual Regret in Cooperative Stochastic Multi-Armed Bandits (NeurIPS 2025 · similarity 0.90)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.