Finite-Memory Near-Optimal Learning for Markov Decision Processes with Long-Run Average Reward.
Jan KretínskýFabian MichelLukas MichelGuillermo A. PérezPublished in: UAI (2020)
Keyphrases
- average reward
- long run
- markov decision processes
- stochastic games
- optimal policy
- reinforcement learning
- state and action spaces
- actor critic
- semi markov decision processes
- hierarchical reinforcement learning
- average cost
- infinite horizon
- partially observable
- state space
- policy iteration
- policy gradient
- finite state
- learning algorithm
- discounted reward
- total reward
- expected cost
- finite horizon
- optimality criterion
- real time dynamic programming
- model free
- decision theoretic planning
- supervised learning
- learning tasks
- control policy
- dynamic programming
- reward function
- multistage
- finite number