Aggregating Optimistic Planning Trees for Solving Markov Decision Processes.
Gunnar KedenburgRaphaël FonteneauRémi MunosPublished in: NIPS (2013)
Keyphrases
- markov decision processes
- transition matrices
- semi markov decision processes
- markov decision problems
- macro actions
- decision theoretic planning
- partially observable
- planning under uncertainty
- solving planning problems
- planning problems
- optimal policy
- state space
- finite state
- policy iteration
- reinforcement learning
- stochastic shortest path
- dynamic programming
- reinforcement learning algorithms
- action sets
- decision processes
- average cost
- action space
- reachability analysis
- finite horizon
- average reward
- infinite horizon
- factored mdps
- heuristic search
- probabilistic planning
- risk sensitive
- linear programming
- state and action spaces
- ai planning
- reward function
- model based reinforcement learning
- search algorithm
- initial state