Offline RL Without Off-Policy Evaluation.
David BrandfonbrenerWilliam F. WhitneyRajesh RanganathJoan BrunaPublished in: CoRR (2021)
Keyphrases
- policy evaluation
- reinforcement learning
- least squares
- model free
- temporal difference
- monte carlo
- policy iteration
- markov decision processes
- variance reduction
- function approximation
- optimal policy
- semi parametric
- td learning
- reinforcement learning algorithms
- partially observable markov decision processes
- state space
- partially observable
- evaluation function
- sample size
- markov decision problems
- optical flow