Nonuniqueness versus Uniqueness of Optimal Policies in Convex Discounted Markov Decision Processes.
Raúl Montes-de-OcaEnrique Lemus-RodríguezFrancisco Salem-SilvaPublished in: J. Appl. Math. (2013)
Keyphrases
- markov decision processes
- optimal policy
- sufficient conditions
- infinite horizon
- state space
- finite horizon
- dynamic programming
- decision problems
- policy iteration
- average reward
- finite state
- reinforcement learning
- long run
- multistage
- partially observable
- average cost
- state dependent
- markov decision process
- reward function
- control policies
- total reward
- reinforcement learning algorithms
- initial state
- discount factor
- decision processes
- policy evaluation
- semi markov decision processes