Where is the Grass Greener? Revisiting Generalized Policy Iteration for Offline Reinforcement Learning.
Lionel BlondéAlexandros KalousisPublished in: CoRR (2021)
Keyphrases
- policy iteration
- reinforcement learning
- markov decision processes
- model free
- policy evaluation
- temporal difference
- optimal policy
- markov decision process
- sample path
- stochastic approximation
- function approximation
- fixed point
- least squares
- average reward
- approximate dynamic programming
- reinforcement learning algorithms
- actor critic
- infinite horizon
- state space
- optimal control
- temporal difference learning
- finite state
- multi agent
- markov decision problems
- control problems
- discounted reward
- average cost
- action selection
- dynamic programming
- optical flow
- long run
- markov chain
- decision making
- learning algorithm