Using Linear Programming for Bayesian Exploration in Markov Decision Processes.
Pablo Samuel CastroDoina PrecupPublished in: IJCAI (2007)
Keyphrases
- markov decision processes
- linear programming
- model based reinforcement learning
- dynamic programming
- policy iteration
- state space
- optimal policy
- interval estimation
- transition matrices
- finite state
- average cost
- linear program
- reinforcement learning
- reinforcement learning algorithms
- average reward
- objective function
- finite horizon
- reachability analysis
- np hard
- decision processes
- bayesian networks
- infinite horizon
- partially observable
- action sets
- factored mdps
- decision theoretic planning
- risk sensitive
- planning under uncertainty
- decision theory
- action space
- markov decision process
- posterior probability
- markov decision problems
- state abstraction
- state and action spaces
- semi markov decision processes
- optimal solution
- optimal control
- integer programming
- reward function
- multistage
- learning algorithm
- utility function