Off-Policy Evaluation via Adaptive Weighting with Data from Contextual Bandits

TL;DR

This paper introduces an adaptive weighting method to improve off-policy evaluation from data collected via contextual bandits, reducing variance and bias in estimators, and enabling reliable confidence intervals.

Contribution

It proposes a novel adaptive weighting approach for doubly robust estimators that controls variance and enhances accuracy in off-policy evaluation from contextual bandit data.

Findings

The improved estimator has lower bias and variance compared to traditional methods.

The t-statistic based on the estimator is asymptotically normal, enabling hypothesis testing.

Empirical results show enhanced accuracy on synthetic and benchmark datasets.

Abstract

It has become increasingly common for data to be collected adaptively, for example using contextual bandits. Historical data of this type can be used to evaluate other treatment assignment policies to guide future innovation or experiments. However, policy evaluation is challenging if the target policy differs from the one used to collect data, and popular estimators, including doubly robust (DR) estimators, can be plagued by bias, excessive variance, or both. In particular, when the pattern of treatment assignment in the collected data looks little like the pattern generated by the policy to be evaluated, the importance weights used in DR estimators explode, leading to excessive variance. In this paper, we improve the DR estimator by adaptively weighting observations to control its variance. We show that a t-statistic based on our improved estimator is asymptotically normal under certain conditions, allowing us to form confidence intervals and test hypotheses. Using synthetic data and public benchmarks, we provide empirical evidence for our estimator's improved accuracy and inferential properties relative to existing alternatives.