Span-Based Optimal Sample Complexity for Average Reward MDPs.
Matthew ZurekYudong ChenPublished in: CoRR (2023)
Keyphrases
- average reward
- sample complexity
- markov decision processes
- optimal policy
- long run
- discounted reward
- semi markov decision processes
- optimality criterion
- reinforcement learning
- policy iteration
- theoretical analysis
- dynamic programming
- learning problems
- model free
- special case
- markov chain
- lower bound
- finite state
- state and action spaces
- generalization error
- learning algorithm
- supervised learning
- active learning
- reward function
- state space
- average cost
- upper bound
- markov decision process
- decision problems
- search space
- fixed point
- partially observable
- utility function
- sample size
- least squares