Cosine Policy Iteration for Solving Infinite-Horizon Markov Decision Processes.
Juan Frausto SolísElizabeth Santiago D.Jaime Mora-VargasPublished in: MICAI (2009)
Keyphrases
- policy iteration
- markov decision processes
- infinite horizon
- markov decision problems
- transition matrices
- optimal policy
- policy iteration algorithm
- finite horizon
- reinforcement learning
- finite state
- state space
- markov decision process
- partially observable
- sample path
- optimal control
- average cost
- average reward
- dynamic programming
- factored mdps
- single item
- policy evaluation
- reinforcement learning algorithms
- planning under uncertainty
- long run
- sufficient conditions
- discounted reward
- action space
- least squares
- model free
- linear programming
- temporal difference
- actor critic
- state and action spaces
- multistage
- cost function
- machine learning