Confounding-Robust Policy Evaluation in Infinite-Horizon Reinforcement Learning.
Nathan KallusAngela ZhouPublished in: NeurIPS (2020)
Keyphrases
- infinite horizon
- policy evaluation
- policy iteration
- markov decision processes
- reinforcement learning
- optimal policy
- optimal control
- finite horizon
- state space
- temporal difference
- markov decision process
- partially observable
- model free
- function approximation
- long run
- dynamic programming
- least squares
- markov decision problems
- decision problems
- finite state
- monte carlo
- reinforcement learning algorithms
- search algorithm
- multistage
- average cost
- markov chain
- fixed point
- variance reduction
- average reward
- multi agent
- partially observable markov decision processes
- action space
- function approximators
- optical flow
- learning algorithm