Exact finite approximations of average-cost countable Markov decision processes.
Arie LeizarowitzAdam ShwartzPublished in: Autom. (2008)
Keyphrases
- average cost
- markov decision processes
- state and action spaces
- finite number
- finite state
- stationary policies
- optimal policy
- state space
- finite horizon
- approximate dynamic programming
- dynamic programming
- policy evaluation
- infinite horizon
- policy iteration
- reinforcement learning
- markov decision process
- initial state
- action space
- average reward
- action sets
- long run
- decision theoretic planning
- partially observable
- reinforcement learning algorithms
- planning under uncertainty
- markov decision chains
- transition matrices
- risk sensitive
- markov decision problems
- control policy
- multistage
- lower bound
- machine learning
- linear program