Optimizing over a Restricted Policy Class in Markov Decision Processes.
Ershad BanijamaliYasin Abbasi-YadkoriMohammad GhavamzadehNikos VlassisPublished in: CoRR (2018)
Keyphrases
- markov decision processes
- optimal policy
- policy iteration
- markov decision process
- infinite horizon
- finite horizon
- average reward
- partially observable
- average cost
- state space
- reinforcement learning
- decision processes
- reward function
- transition matrices
- state and action spaces
- finite state
- action space
- dynamic programming
- decision theoretic planning
- decision problems
- policy evaluation
- model based reinforcement learning
- expected reward
- discounted reward
- reachability analysis
- reinforcement learning algorithms
- continuous state spaces
- long run
- factored mdps
- policy iteration algorithm
- markov decision problems
- state dependent
- partially observable markov decision processes
- action sets
- total reward
- planning under uncertainty
- linear programming
- control policies
- semi markov decision processes
- monte carlo