Offline Policy Selection under Uncertainty
This work is significant for practitioners in reinforcement learning who need to select policies reliably in real-world settings where policy evaluation is uncertain.
This paper addresses the problem of offline policy selection when there is uncertainty in policy evaluation. It proposes BayesDICE, a method for estimating the full belief distribution over policy values, which allows for more flexible policy selection under various downstream metrics. The empirical results show that BayesDICE significantly outperforms existing approaches in policy ranking.
The presence of uncertainty in policy evaluation significantly complicates the process of policy ranking and selection in real-world settings. We formally consider offline policy selection as learning preferences over a set of policy prospects given a fixed experience dataset. While one can select or rank policies based on point estimates of their policy values or high-confidence intervals, access to the full distribution over one's belief of the policy value enables more flexible selection algorithms under a wider range of downstream evaluation metrics. We propose BayesDICE for estimating this belief distribution in terms of posteriors of distribution correction ratios derived from stochastic constraints (as opposed to explicit likelihood, which is not available). Empirically, BayesDICE is highly competitive to existing state-of-the-art approaches in confidence interval estimation. More importantly, we show how the belief distribution estimated by BayesDICE may be used to rank policies with respect to any arbitrary downstream policy selection metric, and we empirically demonstrate that this selection procedure significantly outperforms existing approaches, such as ranking policies according to mean or high-confidence lower bound value estimates.