Research map / Bandits and online learning / Multi-armed bandits
Online Multi-Armed Bandits with Adaptive Inference
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 |
|---|---|---|---|
| replace sample-mean Thompson sampling estimate with adaptively weighted doubly robust estimator (DATS) | TS-Normal · Synthetic domains, K=5..50 | average regret | Helped |
| replace sample-mean Thompson sampling estimate with DATS estimator | TS-Normal · Synthetic domains, K=5..50 | stopping time for best-arm identification | Mixed |
| DATS vs TS and UCB on A/B-test-based bandit | TS-Normal / UCB-Normal · Semi-synthetic A/B test, K=6 | cumulative regret | Helped |
| use unbiased IPW estimator instead of sample mean in TS (no adaptive weights) | TS-Normal · Semi-synthetic A/B test, K=6 | cumulative regret | Hurt |
| use plain doubly robust estimator instead of sample mean in TS | TS-Normal · Semi-synthetic A/B test, K=6 | cumulative regret | Hurt |
| add variance-stabilizing adaptive weights to doubly robust estimator (DATS vs TS-DR) (ablation) | TS-DR · Semi-synthetic A/B test, K=6 | cumulative regret | Helped |
| use propensity clipping instead of arm elimination plus uniform exploration (ablation) | DATS · Synthetic domains | regret and stopping time | No clear effect |
| vary uniform-exploration gamma (0.01, 0.05, 0.1) for DATS (ablation) | DATS · Semi-synthetic low SNR | cumulative regret | No clear effect |
| tune gamma for TS-IPW and TS-DR (ablation) | TS-IPW / TS-DR · Semi-synthetic low SNR | cumulative regret | Helped |
Closest papers by meaning
- Statistical Inference on Multi-armed Bandits with Delayed Feedback (ICML 2023 · similarity 0.91)
- Are sample means in multi-armed bandits positively or negatively biased? (NeurIPS 2019 · similarity 0.90)
- The Choice of Noninformative Priors for Thompson Sampling in Multiparameter Bandit Models (AAAI 2024 · similarity 0.89)
- Robustly Improving Bandit Algorithms with Confounded and Selection Biased Offline Data: A Causal Approach (AAAI 2024 · similarity 0.88)
- Only Pay for What Is Uncertain: Variance-Adaptive Thompson Sampling (ICLR 2024 · 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.