Local Roots: A Tree-Based Subgoal Discovery Method to Accelerate Reinforcement Learning.
Alper DemirErkin ÇildenFaruk PolatPublished in: ECML/PKDD (2) (2016)
Keyphrases
- high accuracy
- significant improvement
- experimental evaluation
- reinforcement learning
- computational complexity
- computationally efficient
- objective function
- classification method
- optimization algorithm
- state space
- model free
- partial order
- fully automatic
- high precision
- clustering method
- model selection
- feature set
- data sets
- knowledge discovery
- probabilistic model
- dynamic programming
- prior knowledge
- similarity measure
- data mining