Online Markov decision processes with policy iteration.
Yao MaHao ZhangMasashi SugiyamaPublished in: CoRR (2015)
Keyphrases
- markov decision processes
- policy iteration
- optimal policy
- finite state
- reinforcement learning
- state space
- dynamic programming
- transition matrices
- average reward
- least squares
- sample path
- fixed point
- factored mdps
- policy evaluation
- markov decision problems
- markov decision process
- average cost
- model free
- planning under uncertainty
- reinforcement learning algorithms
- infinite horizon
- approximate dynamic programming
- decision processes
- finite horizon
- partially observable
- temporal difference
- state and action spaces
- discounted reward
- policy iteration algorithm
- action space
- decision problems
- reward function
- actor critic
- multistage
- sufficient conditions
- probability distribution