Confidence Interval for Off-Policy Evaluation from Dependent Samples via Bandit Algorithm: Approach from Standardized Martingales
This addresses the issue of accurate policy evaluation in reinforcement learning when data is non-i.i.d., which is crucial for practitioners using bandit algorithms, though it is incremental as it builds on existing OPE methods.
This study tackles the problem of off-policy evaluation from dependent samples generated by bandit algorithms, which violate the i.i.d. assumption used in existing methods, by constructing an estimator from a standardized martingale difference sequence. The proposed estimator achieves asymptotic normality without restricting behavior policies and performs better than existing methods in experiments.
This study addresses the problem of off-policy evaluation (OPE) from dependent samples obtained via the bandit algorithm. The goal of OPE is to evaluate a new policy using historical data obtained from behavior policies generated by the bandit algorithm. Because the bandit algorithm updates the policy based on past observations, the samples are not independent and identically distributed (i.i.d.). However, several existing methods for OPE do not take this issue into account and are based on the assumption that samples are i.i.d. In this study, we address this problem by constructing an estimator from a standardized martingale difference sequence. To standardize the sequence, we consider using evaluation data or sample splitting with a two-step estimation. This technique produces an estimator with asymptotic normality without restricting a class of behavior policies. In an experiment, the proposed estimator performs better than existing methods, which assume that the behavior policy converges to a time-invariant policy.