An Algorithm to Identify and Compute Average Optimal Policies in Multichain Markov Decision Processes.
Arie LeizarowitzPublished in: Math. Oper. Res. (2003)
Keyphrases
- markov decision processes
- optimal policy
- average reward
- dynamic programming
- policy iteration
- model based reinforcement learning
- average cost
- reinforcement learning
- finite state
- state space
- real time dynamic programming
- finite horizon
- total reward
- discounted reward
- long run
- sample path
- markov decision process
- discount factor
- reinforcement learning algorithms
- infinite horizon
- monte carlo
- multistage
- learning algorithm
- partially observable
- optimal solution
- state abstraction
- convergence rate
- sufficient conditions
- search space
- optimality criterion
- cost function
- state and action spaces
- lost sales
- multi agent
- computational complexity
- data mining
- np hard
- control policy
- linear programming
- model free