Research map / Bandits and online learning / Multi-armed bandits
Leveraging Observations in Bandits: Between Risks and Benefits
Andrei Lupu, Audrey Durand, Doina Precup
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Learn with Target-UCB from a UCB target instead of standalone UCB | Target-UCB · 2-actions bandit, mu*=0.9 | cumulative regret | Improved |
| Use greedy follower of the UCB target instead of Target-UCB | learner following a UCB target · 2-actions bandit, mu*=0.9 | cumulative regret | Improved |
| Increase alpha of the alpha-optimal target (ablation) | Target-UCB · 2-actions bandit, mu*=0.9, delta_a=0.1 | cumulative regret | Improved |
| Replace greedy follower with Target-UCB when target is poor (low alpha) | learner following alpha-optimal target · 2-actions bandit, alpha=0.001 | cumulative regret | Improved |
| Target-UCB clique of 11 agents vs single UCB | Target-UCB clique · 2-actions bandit, mu*=0.5, delta_a=0.1 | cumulative regret | Improved |
| Target-UCB clique of 11 agents vs UCB clique sharing full information | Target-UCB clique · 2-actions bandit, mu*=0.5 | cumulative regret | Improved |
| Target-UCB graphs of 20 agents vs single UCB | Target-UCB graphs · random 10-actions bandits | cumulative regret | Improved |
| Vary graph structure (chain, loop, clique, random, small-world) (ablation) | Target-UCB graphs · random 10-actions bandits | cumulative regret | No clear effect |
| Target-UCB graphs vs UCB clique of 20 agents | Target-UCB graphs · random 10-actions bandits | cumulative regret | Mixed |
| Target-UCB learning from a single human target | Target-UCB · human 2-actions bandit, mu*=0.6 | cumulative regret | Improved |
| Clique of four Target-UCB agents vs clique of four humans | Target-UCB clique · human 2-actions bandit, mu*=0.6 | cumulative regret | Improved |
Closest papers by meaning
- Causal Bandits: The Pareto Optimal Frontier of Adaptivity, a Reduction to Linear Bandits, and Limitations around Unknown Marginals (ICML 2024 · similarity 0.90)
- Robustly Improving Bandit Algorithms with Confounded and Selection Biased Offline Data: A Causal Approach (AAAI 2024 · similarity 0.89)
- Bandits with Abstention under Expert Advice (NeurIPS 2024 · similarity 0.89)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.89)
- Multi-Agent Imitation Learning: Value is Easy, Regret is Hard (NeurIPS 2024 · similarity 0.88)
- Block Optimism for Nonstationary Bandits with Latent Linear Dynamics (NeurIPS 2026 · similarity 0.88)
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.
