Off-policy Evaluation in Infinite-Horizon Reinforcement Learning with Latent Confounders.
Andrew BennettNathan KallusLihong LiAli MousaviPublished in: CoRR (2020)
Keyphrases
- policy evaluation
- infinite horizon
- policy iteration
- markov decision processes
- reinforcement learning
- optimal policy
- optimal control
- finite horizon
- state space
- markov decision process
- temporal difference
- dynamic programming
- long run
- partially observable markov decision processes
- model free
- markov decision problems
- finite state
- partially observable
- decision problems
- function approximation
- latent variables
- least squares
- reinforcement learning algorithms
- multistage
- average reward
- lost sales
- decision making
- learning algorithm
- gaussian process
- lead time
- belief state
- supervised learning
- inventory level
- objective function