On-Line Model-Based Continuous State Reinforcement Learning Using Background Knowledge.
Bernhard HengstPublished in: Australasian Conference on Artificial Intelligence (2012)
Keyphrases
- background knowledge
- continuous state
- reinforcement learning
- model free
- policy search
- inductive logic programming
- robot navigation
- domain knowledge
- action space
- continuous state and action spaces
- logic programs
- control policies
- function approximation
- finite state
- theory revision
- prior knowledge
- state dependent
- state space
- planning problems
- reinforcement learning algorithms
- partially observable markov decision processes
- learning algorithm
- knowledge base
- machine learning
- search engine
- state action
- transfer learning
- semantic information
- action selection
- learning process
- markov decision processes
- temporal difference
- training data
- pairwise
- hidden state
- markov chain
- supervised learning
- dynamic programming