Research map / Bandits and online learning / Multi-armed bandits
Constrained Best Arm Identification with Tests for Feasibility
Ting Cai, Kirthevasan Kandasamy
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 |
|---|---|---|---|
| test feasibility constraints one at a time (SAMPLE-FEASIBILITY) instead of sampling all constraints together (ablation) | feasible best arm identification · synthetic and drug discovery experiments | sample complexity | Improved |
| balance performance and feasibility testing instead of feasibility-first | Algorithm 1 vs F-first · Experiment 1 instance (b) | relative number of samples: 3.13 -> 1.0 | Improved |
| balance performance and feasibility testing instead of performance-first | Algorithm 1 vs P-first · Experiment 1 instance (a) | relative number of samples: 3.09 -> 1.0 | Improved |
| use Algorithm 1 instead of the best specialized baseline on its favorable instance | Algorithm 1 vs F-first on instance (a) · Experiment 1 instance (a) | relative number of samples: 0.58 vs 1.0 | Hurt |
| use Algorithm 1 instead of P-first on instance (b) | Algorithm 1 vs P-first · Experiment 1 instance (b) | relative number of samples: 0.84 vs 1.0 | Hurt |
| use Algorithm 1 on mixed instance with several suboptimal arm types | Algorithm 1 vs all baselines · Experiment 1 instance (c) | relative number of samples: best baseline 2.78 -> 1.0 | Improved |
| increase number of constraints N from 2 to 64 | Algorithm 1 · Experiment 2 | sample complexity | Improved |
| apply Algorithm 1 to real drug dosage data | Algorithm 1 · Drug discovery (Genovese et al. 2013) | sample complexity | Improved |
| use a tighter confidence bound only for TF-LUCB-C (ablation) | TF-LUCB-C · Appendix confidence bound comparison | sample complexity | No clear effect |
Closest papers by meaning
- Optimal Top-Two Method for Best Arm Identification and Fluid Analysis (NeurIPS 2024 · similarity 0.88)
- Balancing Performance and Costs in Best Arm Identification (NeurIPS 2025 · similarity 0.86)
- Optimal Best Arm Identification under Differential Privacy (NeurIPS 2025 · similarity 0.86)
- Optimal Multi-Fidelity Best-Arm Identification (NeurIPS 2024 · similarity 0.86)
- Optimal Batched Best Arm Identification (NeurIPS 2024 · similarity 0.85)
- Quantum Best Arm Identification with Quantum Oracles (AAAI 2025 · similarity 0.85)
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.
