Minimax Value Interval for Off-Policy Evaluation and Policy Optimization.
Nan JiangJiawei HuangPublished in: NeurIPS (2020)
Keyphrases
- policy evaluation
- least squares
- temporal difference
- monte carlo
- markov decision processes
- policy iteration
- reinforcement learning
- model free
- variance reduction
- optimal policy
- function approximation
- evaluation function
- semi parametric
- partially observable markov decision processes
- constrained optimization
- dynamic programming
- step size
- reinforcement learning algorithms
- finite state
- markov decision problems
- policy gradient
- worst case