Weighted difference approximation of value functions for slow-discounting Markov Decision Processes.
Yinlam ChowJunjie QinPublished in: CDC (2014)
Keyphrases
- markov decision processes
- convex functions
- finite state
- optimal policy
- decision diagrams
- policy iteration
- reinforcement learning
- state space
- transition matrices
- decision theoretic planning
- dynamic programming
- policy iteration algorithm
- reachability analysis
- planning under uncertainty
- reinforcement learning algorithms
- factored mdps
- model based reinforcement learning
- decision processes
- average reward
- semi markov decision processes
- risk sensitive
- policy evaluation
- action space
- infinite horizon
- finite horizon
- markov decision process
- average cost
- partially observable
- linear program
- learning algorithm
- action sets
- reward function
- least squares