Unbounded cost Markov decision processes with limsup and liminf average criteria: new conditions.
Quanxin ZhuXianping GuoYonglong DaiPublished in: Math. Methods Oper. Res. (2005)
Keyphrases
- markov decision processes
- average cost
- optimal policy
- stationary policies
- finite state
- state space
- infinite horizon
- transition matrices
- finite horizon
- dynamic programming
- initial state
- reinforcement learning
- policy iteration
- long run
- factored mdps
- risk sensitive
- average reward
- markov decision process
- reachability analysis
- optimal control
- finite number
- decision processes
- decision theoretic planning
- sufficient conditions
- partially observable
- discounted reward
- state and action spaces
- model based reinforcement learning
- total cost
- optimality criterion
- planning under uncertainty
- reward function
- linear programming
- reinforcement learning algorithms
- linear program
- multistage
- decision diagrams
- search algorithm