Conditions for the uniqueness of optimal policies of discounted Markov decision processes.
Daniel Cruz-SuárezRaúl Montes-de-OcaFrancisco Salem-SilvaPublished in: Math. Methods Oper. Res. (2004)
Keyphrases
- markov decision processes
- optimal policy
- sufficient conditions
- infinite horizon
- state space
- average reward
- finite state
- finite horizon
- decision problems
- policy iteration
- average cost
- dynamic programming
- state dependent
- reinforcement learning
- markov decision process
- stationary policies
- multistage
- discounted reward
- semi markov decision processes
- decision processes
- partially observable
- markov decision problems
- long run
- action space
- state abstraction
- discount factor
- lost sales
- control policies
- reward function
- state and action spaces
- total reward
- reinforcement learning algorithms
- partially observable markov decision processes