Doubly Robust Policy Evaluation and Optimization

Abstract

We study sequential decision making in environments where rewards are only partially observed, but can be modeled as a function of observed contexts and the chosen action by the decision maker. This setting, known as contextual bandits, encompasses a wide variety of applications such as health care, content recommendation and Internet advertising. A central task is evaluation of a new policy given historic data consisting of contexts, actions and received rewards. The key challenge is that the past data typically does not faithfully represent proportions of actions taken by a new policy. Previous approaches rely either on models of rewards or models of the past policy. The former are plagued by a large bias whereas the latter have a large variance. In this work, we leverage the strengths and overcome the weaknesses of the two approaches by applying the doubly robust estimation technique to the problems of policy evaluation and optimization. We prove that this approach yields accurate value estimates when we have either a good (but not necessarily consistent) model of rewards or a good (but not necessarily consistent) model of past policy. Extensive empirical comparison demonstrates that the doubly robust estimation uniformly improves over existing techniques, achieving both lower variance in value estimation and better policies. As such, we expect the doubly robust approach to become common practice in policy evaluation and optimization.

Published in at http://dx.doi.org/10.1214/14-STS500 the Statistical Science (http://www.imstat.org/sts/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (6)

• A Contextual-Bandit Approach to Personalized News Article Recommendation
• Demystifying Double Robustness: A Comparison of Alternative Strategies for Estimating a Population Mean from Incomplete Data
• Unbiased Offline Evaluation of Contextual-bandit-based News Article Recommendation Algorithms
• Doubly Robust Policy Evaluation and Learning
• Contextual Bandit Algorithms with Supervised Learning Guarantees
• Sample-efficient Nonstationary Policy Evaluation for Contextual Bandits

Cited by in corpus (18)

• A Survey on Causal Inference for Recommendation
• Doubly Robust Off-Policy Evaluation for Ranking Policies under the Cascade Behavior Model
• Offline Contextual Multi-armed Bandits for Mobile Health Interventions: A Case Study on Emotion Regulation
• Causal mediation analysis for stochastic interventions
• Evaluating the Robustness of Off-Policy Evaluation
• Doubly-Robust Estimation for Correcting Position-Bias in Click Feedback for Unbiased Learning to Rank
• Causal Etiology of the Research of James M. Robins
• Optimal Baseline Corrections for Off-Policy Contextual Bandits
• PAC-Bayesian Lifelong Learning For Multi-Armed Bandits
• Off-Policy Evaluation of Slate Bandit Policies via Optimizing Abstraction
• Debiased Off-Policy Evaluation for Recommendation Systems
• Selecting and ranking individualized treatment rules with unmeasured confounding
• PAC-Bayes Bounds for Bandit Problems: A Survey and Experimental Comparison
• Towards Robust Off-Policy Evaluation via Human Inputs
• On the Reliability of Sampling Strategies in Offline Recommender Evaluation
• Learning Action Embeddings for Off-Policy Evaluation
• Meta Off-Policy Estimation
• A Statistical Test for the Benefits of Personalizing Interventions