Offline RL Without Off-Policy Evaluation.
David BrandfonbrenerWill WhitneyRajesh RanganathJoan BrunaPublished in: NeurIPS (2021)
Keyphrases
- policy evaluation
- reinforcement learning
- least squares
- temporal difference
- model free
- markov decision processes
- monte carlo
- policy iteration
- function approximation
- variance reduction
- semi parametric
- optimal policy
- td learning
- reinforcement learning algorithms
- step size
- fixed point
- neural network
- linear model
- partially observable markov decision processes
- statistical inference
- evaluation function
- state space
- decision making
- dynamical systems
- model selection
- supervised learning
- search space
- support vector