Nearly optimal policies in risk-sensitive positive dynamic programming on discrete spaces.
Rolando Cavazos-CadenaRaúl Montes-de-OcaPublished in: Math. Methods Oper. Res. (2000)
Keyphrases
- risk sensitive
- optimal policy
- dynamic programming
- markov decision processes
- markov decision problems
- optimal control
- control policies
- average cost
- state space
- infinite horizon
- finite horizon
- decision problems
- finite state
- reinforcement learning
- long run
- multistage
- dynamic programming algorithms
- average reward
- policy iteration
- markov decision process
- reward function
- linear program
- initial state
- decision processes
- linear programming
- partially observable
- reinforcement learning algorithms
- sufficient conditions
- data mining