Towards Interpretable Reinforcement Learning with State Abstraction Driven by External Knowledge.
Nicolas BougieRyutaro IchisePublished in: IEICE Trans. Inf. Syst. (2020)
Keyphrases
- state abstraction
- external knowledge
- reinforcement learning
- state space
- domain knowledge
- hierarchical reinforcement learning
- wordnet
- markov decision processes
- text categorization
- path finding
- reinforcement learning algorithms
- function approximation
- knowledge sources
- optimal policy
- document clustering
- textual information
- temporal difference
- bag of words
- machine learning
- model free
- initial state
- semantic features
- knowledge structures
- web directories
- supervised learning
- search algorithm
- knowledge base