Best Arm Identification with Possibly Biased Offline Data

TL;DR

This paper addresses the challenge of best arm identification using biased offline data, proposing an adaptive algorithm that balances offline and online data, with theoretical guarantees and empirical validation.

Contribution

Introduces LUCB-H, an adaptive algorithm for BAI with biased offline data, providing theoretical analysis and demonstrating improved performance over standard methods.

Findings

LUCB-H matches standard LUCB when offline data is misleading.

LUCB-H outperforms standard LUCB when offline data is helpful.

Numerical experiments confirm LUCB-H's robustness and adaptability.

Abstract

We study the best arm identification (BAI) problem with potentially biased offline data in the fixed confidence setting, which commonly arises in real-world scenarios such as clinical trials. We prove an impossibility result for adaptive algorithms without prior knowledge of the bias bound between online and offline distributions. To address this, we propose the LUCB-H algorithm, which introduces adaptive confidence bounds by incorporating an auxiliary bias correction to balance offline and online data within the LUCB framework. Theoretical analysis shows that LUCB-H matches the sample complexity of standard LUCB when offline data is misleading and significantly outperforms it when offline data is helpful. We also derive an instance-dependent lower bound that matches the upper bound of LUCB-H in certain scenarios. Numerical experiments further demonstrate the robustness and adaptability of LUCB-H in effectively incorporating offline data.