A Provably-Efficient Model-Free Algorithm for Infinite-Horizon Average-Reward Constrained Markov Decision Processes.
Honghao WeiXin LiuLei YingPublished in: AAAI (2022)
Keyphrases
- average reward
- markov decision processes
- policy iteration
- model free
- infinite horizon
- optimal policy
- long run
- reinforcement learning
- total reward
- dynamic programming
- stochastic games
- policy evaluation
- optimality criterion
- semi markov decision processes
- state space
- finite horizon
- discounted reward
- reinforcement learning algorithms
- discount factor
- finite state
- average cost
- partially observable
- function approximation
- decision problems
- planning under uncertainty
- fixed point
- markov decision process
- temporal difference
- action space
- markov decision problems
- learning algorithm
- multi agent
- search space
- policy gradient
- optimal control