Abstraction in Model Based Partially Observable Reinforcement Learning Using Extended Sequence Trees.
Erkin ÇildenFaruk PolatPublished in: IAT (2012)
Keyphrases
- partially observable
- reinforcement learning
- hidden state
- model free
- state space
- markov decision processes
- partial observability
- decision problems
- dynamical systems
- partially observable environments
- partially observable domains
- markov decision problems
- action models
- infinite horizon
- optimal policy
- function approximation
- learning algorithm
- reinforcement learning algorithms
- belief space
- multi agent
- belief state
- reward function
- temporal difference
- partially observable markov decision processes
- special case
- partial observations
- machine learning
- policy iteration
- action selection
- long run
- dynamic programming
- inverse reinforcement learning