Offline Policy Selection under Uncertainty
arXiv:2012.06919
Abstract
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.
References in corpus (10)
- Doubly Robust Policy Evaluation and Learning
- Deep Exploration via Randomized Value Functions
- Breaking the Curse of Horizon: Infinite-Horizon Off-Policy Estimation
- Doubly Robust Off-policy Value Evaluation for Reinforcement Learning
- Provable Bayesian Inference via Particle Mirror Descent
- Batch Value-function Approximation with Only Realizability
- Reinforcement Learning via Fenchel-Rockafellar Duality
- Statistical Bootstrapping for Uncertainty Estimation in Off-Policy Evaluation
- Confident Off-Policy Evaluation and Selection through Self-Normalized Importance Weighting
- A Linearly Relaxed Approximate Linear Program for Markov Decision Processes