Off-policy Bandits with Deficient Support

TL;DR

This paper addresses the challenge of off-policy learning in contextual bandits with support-deficient data, proposing three approaches to ensure reliable policy learning despite limited support in logged data.

Contribution

It introduces the first systematic analysis of support deficiency in contextual-bandit learning and proposes three methods to mitigate its impact.

Findings

Restricting the action space improves learning in support-deficient data.

Reward extrapolation can compensate for missing support under certain conditions.

Restricting the policy space offers theoretical guarantees despite support limitations.

Abstract

Learning effective contextual-bandit policies from past actions of a deployed system is highly desirable in many settings (e.g. voice assistants, recommendation, search), since it enables the reuse of large amounts of log data. State-of-the-art methods for such off-policy learning, however, are based on inverse propensity score (IPS) weighting. A key theoretical requirement of IPS weighting is that the policy that logged the data has "full support", which typically translates into requiring non-zero probability for any action in any context. Unfortunately, many real-world systems produce support deficient data, especially when the action space is large, and we show how existing methods can fail catastrophically. To overcome this gap between theory and applications, we identify three approaches that provide various guarantees for IPS-based learning despite the inherent limitations of support-deficient data: restricting the action space, reward extrapolation, and restricting the policy space. We systematically analyze the statistical and computational properties of these three approaches, and we empirically evaluate their effectiveness. In addition to providing the first systematic analysis of support-deficiency in contextual-bandit learning, we conclude with recommendations that provide practical guidance.