A Stopping Rule for Discounted Markov Decision Processes with Finite Action Sets.
Raúl Montes-de-OcaDaniel Cruz-SuárezEnrique Lemus-RodríguezPublished in: Kybernetika (2009)
Keyphrases
- markov decision processes
- action sets
- stationary policies
- state and action spaces
- optimal policy
- finite state
- reinforcement learning
- state space
- average cost
- dynamic programming
- finite horizon
- policy iteration
- transition matrices
- average reward
- markov decision process
- infinite horizon
- action space
- partially observable
- reinforcement learning algorithms
- decision processes
- decentralized control
- finite number
- sufficient conditions
- total reward
- discounted reward
- learning algorithm