New inference strategies for solving Markov Decision Processes using reversible jump MCMC
Matthias HoffmanHendrik KückNando de FreitasArnaud DoucetPublished in: CoRR (2012)
Keyphrases
- markov decision processes
- transition matrices
- reversible jump mcmc
- semi markov decision processes
- bayesian networks
- state space
- optimal policy
- finite state
- decision theoretic planning
- dynamic programming
- stochastic shortest path
- reinforcement learning
- infinite horizon
- average reward
- model based reinforcement learning
- factored mdps
- risk sensitive
- markov decision problems
- planning under uncertainty
- reinforcement learning algorithms
- action sets
- reachability analysis
- simulated annealing
- policy iteration
- finite horizon
- partially observable
- decision diagrams
- average cost
- interval estimation
- decision processes
- state and action spaces
- neural network
- action space
- learning algorithm