New inference strategies for solving Markov Decision Processes using reversible jump MCMC.
Matthias HoffmanHendrik KückNando de FreitasArnaud DoucetPublished in: UAI (2009)
Keyphrases
- markov decision processes
- transition matrices
- semi markov decision processes
- reversible jump mcmc
- bayesian networks
- optimal policy
- dynamic programming
- state space
- reinforcement learning
- finite state
- markov decision problems
- stochastic shortest path
- risk sensitive
- policy iteration
- partially observable
- reinforcement learning algorithms
- state and action spaces
- factored mdps
- decision theoretic planning
- average reward
- average cost
- infinite horizon
- finite horizon
- markov decision process
- reachability analysis
- model based reinforcement learning
- decision processes
- planning under uncertainty
- action sets
- simulated annealing
- decision problems
- action space
- sufficient conditions
- real time dynamic programming
- multi agent