Optimization of parametric policies of Markov decision processes under a variance criterion.
Li XiaPublished in: WODES (2016)
Keyphrases
- markov decision processes
- optimal policy
- markov decision process
- decision processes
- state space
- reward function
- dynamic programming
- finite state
- average cost
- transition matrices
- decentralized control
- policy iteration
- total reward
- reinforcement learning
- stationary policies
- macro actions
- finite horizon
- average reward
- expected reward
- partially observable
- discounted reward
- decision theoretic planning
- action space
- decision problems
- markov decision problems
- factored mdps
- policy iteration algorithm
- model based reinforcement learning
- planning under uncertainty
- infinite horizon
- reinforcement learning algorithms
- control policies
- risk sensitive
- long run
- optimality criterion
- multistage
- sufficient conditions
- partially observable markov decision processes
- state abstraction
- decision diagrams
- state and action spaces
- action sets
- initial state
- markov chain
- reachability analysis
- semi markov decision processes
- model checking