Trajectory Sampling Value Iteration: Improved Dyna Search for MDPs.
Yicheng ZhouQuan LiuQi-ming FuZongzhang ZhangPublished in: AAMAS (2015)
Keyphrases
- markov decision processes
- state space
- search algorithm
- search space
- optimal policy
- reinforcement learning
- policy iteration
- dynamic programming
- markov decision process
- heuristic search
- sample size
- search strategies
- factored mdps
- branch and bound
- decision theoretic planning
- average reward
- tree search algorithm
- stochastic shortest path
- finite horizon
- trajectory data
- random sampling
- machine learning
- image sequences
- information retrieval