Approximate Linear-Programming Algorithms for Graph-Based Markov Decision Processes.
Nicklas ForsellRégis SabbadinPublished in: ECAI (2006)
Keyphrases
- markov decision processes
- policy iteration
- factored mdps
- linear programming
- optimal policy
- reinforcement learning
- policy evaluation
- dynamic programming
- infinite horizon
- transition matrices
- computational complexity
- state space
- average cost
- decision problems
- stochastic shortest path
- reachability analysis
- finite horizon
- partially observable markov decision processes
- reinforcement learning algorithms
- finite state
- linear program
- approximate solutions
- multistage
- action space
- learning algorithm
- decision theoretic planning
- discounted reward
- convergence rate
- multi agent