Learning the Difference between Partially Observable Dynamical Systems.
Sami ZhiouaDoina PrecupFrançois LavioletteJosée DesharnaisPublished in: ECML/PKDD (2) (2009)
Keyphrases
- partially observable
- dynamical systems
- reinforcement learning
- action models
- hidden state
- predictive state representations
- partial observations
- state space
- partially observable environments
- markov decision processes
- decision problems
- dynamic systems
- partial observability
- markov decision problems
- supervised learning
- learning tasks
- belief state
- infinite horizon
- optimal policy
- search algorithm
- optimal solution
- decision making