paper-with-me

홈 › Papers

Are sample means in multi-armed bandits positively or negatively biased?

2019-05-27 · NeurIPS 2019 12 · Jaehyeok Shin, Aaditya Ramdas, Alessandro Rinaldo

It is well known that in stochastic multi-armed bandits (MAB), the sample mean of an arm is typically not an unbiased estimator of its true mean. In this paper, we decouple three different sources of this selection bias: adaptive \emph{sampling} of arms, adaptive \emph{stopping} of the experiment, and adaptively \emph{choosing} which arm to study. Through a new notion called ``optimism'' that captures certain natural monotonic behaviors of algorithms, we provide a clean and unified analysis of how optimistic rules affect the sign of the bias. The main takeaway message is that optimistic sampling induces a negative bias, but optimistic stopping and optimistic choosing both induce a positive bias. These results are derived in a general stochastic MAB setup that is entirely agnostic to the final aim of the experiment (regret minimization or best-arm identification or anything else). We provide examples of optimistic rules of each type, demonstrate that simulations confirm our theoretical predictions, and pose some natural but hard open problems.

📄 PDF Abstract BibTeX arXiv:1905.11397

Code (0)

등록된 구현이 없습니다.

Tasks

Multi-Armed BanditsSelection bias

Similar Papers 제목 키워드 기반

On Top-k Selection in Multi-Armed Bandits and Hidden Bipartite Graphs

2015-12-01 · NeurIPS 2015 12 · Wei Cao, Jian Li, Yufei Tao, Zhize Li

This paper discusses how to efficiently choose from $n$ unknowndistributions the $k$ ones whose means are the greatest by a certainmetric, up to a small relative error. We study the topic under twostandard settings---mul…

Multi-Armed Bandits

On conditional versus marginal bias in multi-armed bandits

2020-02-19 · ICML 2020 1 · Jaehyeok Shin, Aaditya Ramdas, Alessandro Rinaldo

The bias of the sample means of the arms in multi-armed bandits is an important issue in adaptive data analysis that has recently received considerable attention in the literature. Existing results relate in precise ways…

Multi-Armed Bandits

Trading off rewards and errors in multi-armed bandits

2026-05-01 · Akram Erraqabi, Alessandro Lazaric, Michal Valko, Emma Brunskill 외 arxiv

In multi-armed bandits, the most-explored arms are the most informative, while reward maximization typically pulls only the best arm. We study the tradeoff between identifying arm means accurately and accumulating reward…

Multi-Armed Bandits

Reward Maximization for Pure Exploration: Minimax Optimal Good Arm Identification for Nonparametric Multi-Armed Bandits

2024-10-21 · Brian Cho, Dominik Meier, Kyra Gan, Nathan Kallus

In multi-armed bandits, the tasks of reward maximization and pure exploration are often at odds with each other. The former focuses on exploiting arms with the highest means, while the latter may require constant explora…

Multi-Armed Banditsvalid

Optimism Stabilizes Thompson Sampling for Adaptive Inference

2026-02-05 · Shunxing Yan, Han Zhong arxiv

Thompson sampling (TS) is widely used for stochastic multi-armed bandits, yet its inferential properties under adaptive data collection are subtle. Classical asymptotic theory for sample means can fail because arm-specif…

Multi-Armed Bandits