Batch Policy Learning in Average Reward Markov Decision Processes
This addresses the challenge of learning optimal policies from historical data in mobile health applications where maximizing long-term average reward is critical.
The authors tackled the problem of batch policy learning in infinite horizon Markov Decision Processes by developing a doubly robust estimator for average reward that achieves semiparametric efficiency, and established finite-sample regret guarantees with simulation and mobile health application results showing competitive performance.
We consider the batch (off-line) policy learning problem in the infinite horizon Markov Decision Process. Motivated by mobile health applications, we focus on learning a policy that maximizes the long-term average reward. We propose a doubly robust estimator for the average reward and show that it achieves semiparametric efficiency. Further we develop an optimization algorithm to compute the optimal policy in a parameterized stochastic policy class. The performance of the estimated policy is measured by the difference between the optimal average reward in the policy class and the average reward of the estimated policy and we establish a finite-sample regret guarantee. The performance of the method is illustrated by simulation studies and an analysis of a mobile health study promoting physical activity.