A new method for discovering subgoals and constructing options in reinforcement learning.
Marzieh DavoodabadiHamid BeigyPublished in: IICAI (2011)
Keyphrases
- dynamic programming
- high accuracy
- high precision
- cost function
- experimental evaluation
- objective function
- fully automatic
- theoretical analysis
- significant improvement
- pairwise
- reinforcement learning
- supervised learning
- computational cost
- optimization algorithm
- synthetic data
- segmentation method
- function approximation
- transition model