Work

Inference for Batched Bandits

Year: 2020

Type: article

Abstract: asymptotically normal on data collected using standard bandit algorithms when there is no unique optimal arm. This asymptotic non-normality result implies that the naive assumption that the OLS estimator is approximately normal can lead to Type-1 error inflation and confidence intervals with below-nominal coverage probabilities. Second, we introduce the Batched OLS estimator (BOLS) that we prove is (1) asymptotically normal on data collected from both multi-arm and contextual bandits and (2) robust to non-stationarity in the baseline reward.

Source: PubMed

DOI: 10.48550/arxiv.2002.03217

Authors: Zhang, Kelly W., Janson, Lucas, Murphy, Susan A.

Institution: Harvard University

Language: English

Topic: Advanced Bandit Algorithms Research

Subfield: Management Science and Operations Research

Field: Decision Sciences

Domain: Social Sciences

Open Access status: green

Funders: National Institute on Alcohol Abuse and Alcoholism, National Institute on Drug Abuse, National Cancer Institute, National Institute of Biomedical Imaging and Bioengineering

Awards: Center for Complex Data to Knowledge in Drug Abuse and HIV Behavioral Science, Novel use of mHealth data to identify states of vulnerability and receptivity to JITAIs, Data-Based Methods for Just-In-Time Adaptive interventions in Alcohol Use, mHealth Center for Discovery, Optimization, and Translation of Temporally-Precise Interventions (mDOT)