Near Optimality of Finite Memory Feedback Policies in Partially Observed Markov Decision Processes.
Ali Devran KaraSerdar YükselPublished in: CoRR (2020)
Keyphrases
- partially observed
- markov decision processes
- stationary policies
- average cost
- expected reward
- optimal policy
- state and action spaces
- average reward
- finite state
- state space
- action sets
- markov decision process
- reinforcement learning
- finite number
- reward function
- dynamic programming
- finite horizon
- reinforcement learning algorithms
- markov decision problems
- total reward
- policy iteration
- transition matrices
- reachability analysis
- infinite horizon
- action space
- decision processes
- linear program
- policy iteration algorithm
- discounted reward
- decentralized control
- decision theoretic planning
- factored mdps
- risk sensitive
- sufficient conditions
- control policies
- planning under uncertainty
- partially observable
- decision diagrams
- semi markov decision processes
- model based reinforcement learning
- long run
- multistage