Double Reinforcement Learning for Efficient Off-Policy Evaluation in Markov Decision Processes.
Nathan KallusMasatoshi UeharaPublished in: J. Mach. Learn. Res. (2020)
Keyphrases
- markov decision processes
- policy evaluation
- reinforcement learning
- policy iteration
- optimal policy
- model free
- state space
- temporal difference
- least squares
- reinforcement learning algorithms
- function approximation
- finite state
- dynamic programming
- monte carlo
- partially observable
- average cost
- decision processes
- planning under uncertainty
- markov decision process
- semi parametric
- partially observable markov decision processes
- action space
- infinite horizon
- optimal control
- state and action spaces
- variance reduction
- state variables
- average reward