Research map / Bandits and online learning / Multi-armed bandits
Box Thirding: Anytime Best Arm Identification under Insufficient Sampling
SEOHWA HWANG, Junyong Park
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 |
|---|---|---|---|
| use B3 (anytime) instead of BSH | anytime BAI · N=128, T=896 | average simple regret: 0.0992 -> 0.0490 | Improved |
| use B3 (anytime) instead of BUCB | anytime BAI · N=128, T=896 | average simple regret: 0.1252 -> 0.0490 | Improved |
| use B3 (anytime) instead of BSH | anytime BAI · N=1,024, T=10,240 | average simple regret: 0.1531 -> 0.0268 | Improved |
| use B3 (anytime) instead of BUCB | anytime BAI · N=1,024, T=10,240 | average simple regret: 0.2558 -> 0.0268 | Improved |
| use B3 (no knowledge of T) instead of UCB-E (knows T) | fixed-budget BAI · N=128, T=896 | average simple regret: 0.0664 -> 0.0490 | Improved |
| use B3 (no knowledge of T) instead of UCB-E (knows T) | fixed-budget BAI · N=1,024, T=10,240 | average simple regret: 0.0403 -> 0.0268 | Improved |
| use B3 instead of SH with r0=1.7 (which knows T) | fixed-budget BAI · N=1,024, T=10,240 | average simple regret: 0.0195 -> 0.0268 | Hurt |
| use B3 instead of SH with r0=1.7 (which knows T) | fixed-budget BAI · N=128, T=896 | average simple regret: 0.0446 -> 0.0490 | Hurt |
| change SH base sampling parameter r0 from 2.0 to 1.7 | SH · N=128, T=896 | average simple regret: 0.0516 -> 0.0446 | Improved |
| change SH base sampling parameter r0 from 2.0 to 1.7 | SH · N=1,024, T=10,240 | average simple regret: 0.0288 -> 0.0195 | Improved |
| increase reward noise from deterministic to high-noise | US vs B3 · NYCCC 893, T=10,000 | average simple regret | Mixed |
| replace Gaussian reward noise with Bernoulli or power-law rewards | B3 · NYCCC 893 | average simple regret | No clear effect |
Closest papers by meaning
- Optimal Batched Best Arm Identification (NeurIPS 2024 · similarity 0.89)
- Balancing Performance and Costs in Best Arm Identification (NeurIPS 2025 · similarity 0.89)
- Optimal Multi-Fidelity Best-Arm Identification (NeurIPS 2024 · similarity 0.89)
- Breaking the $\log(1/\Delta_2)$ Barrier: Better Batched Best Arm Identification with Adaptive Grids (ICLR 2025 · similarity 0.87)
- Fixing the Loose Brake: Exponential-Tailed Stopping Time in Best Arm Identification (ICML 2025 · similarity 0.87)
- Does 1/2-Tsallis-INF Also Work Well for Best-Arm Identification? (NeurIPS 2026 · 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.
