Research map / Bandits and online learning / Multi-armed bandits
Almost Optimal Anytime Algorithm for Batched Multi-Armed Bandits
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 BABA's batched anytime time grid instead of fully sequential play | batched MAB anytime setting · Gaussian and Bernoulli bandits, K=2 and K=5, T up to 10^5 | regret relative to KL-UCB: comparable regret with 10 batches vs 10^5 | No clear effect |
| Use BABA at small horizon (T=2000) instead of KL-UCB | BABA · Gaussian rewards, T=2000 | regret | Hurt |
| Set the epoch growth function f(x)=max{ceil(x^{1+1/(1+ilog_alpha x)}),2x} so batches scale as log log T times ilog_alpha(T) | BABA · theoretical batch complexity | number of batches | Helped |
| Impose a static grid with at most c log log T batches (lower-bound setting) | anytime batched bandit algorithms · theoretical lower bound | achievability of asymptotic optimality | Hurt |
| Regret increases sharply during exploration stages (suboptimal arms pulled) | BABA · Gaussian rewards, K=2 | regret: regret rises from 17.0 to 24.2 between steps 520 and 560 | Hurt |
Closest papers by meaning
- Optimal Batched Linear Bandits (ICML 2024 · similarity 0.90)
- Optimal and Practical Batched Linear Bandit Algorithm (ICML 2025 · similarity 0.90)
- The Batch Complexity of Bandit Pure Exploration (ICML 2025 · similarity 0.88)
- Achieving adaptivity and optimality for multi-armed bandits using Exponential-Kullback Leibler Maillard Sampling (NeurIPS 2026 · similarity 0.88)
- Batched Stochastic Linear Bandits with 1-Bit Communication Constraints (NeurIPS 2026 · similarity 0.88)
- Adaptive Regret for Bandits Made Possible: Two Queries Suffice (ICLR 2024 · similarity 0.87)
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.