On overfitting and asymptotic bias in batch reinforcement learning with partial observability.
Vincent François-LavetDamien ErnstRaphael FonteneauPublished in: CoRR (2017)
Keyphrases
- partial observability
- reinforcement learning
- partially observable
- symbolic model checking
- planning problems
- markov decision process
- state space
- partially observable markov decision processes
- fully observable
- belief state
- function approximation
- dynamical systems
- belief space
- model free
- learning agent
- optimal control
- learning algorithm
- planning under partial observability
- reinforcement learning algorithms
- temporal difference
- action selection
- planning domains
- evaluation function
- decision problems
- markov decision processes
- dynamic programming