Title :
Networks with input gates for situation-dependent input selection in reinforcement learning
Author :
Murata, Junichi ; Suzuki, Masafumi ; Hirasawa, Kotaro
Author_Institution :
Kyushu Univ., Fukuoka, Japan
fDate :
6/24/1905 12:00:00 AM
Abstract :
A method is proposed for situation-dependent input selection and learning acceleration in Q-learning. Q-values are expressed by an RBF network which has an input gate attached to each of its input channels in order to capture, by learning, situation-dependent relevance or usefulness of the input
Keywords :
learning (artificial intelligence); radial basis function networks; Q-learning; Q-values; input gates; learning acceleration; reinforcement learning; situation-dependent input selection; situation-dependent relevance; situation-dependent usefulness; Acceleration; Function approximation; Input variables; Intelligent networks; Learning; Neural networks; Pattern recognition; Radial basis function networks; State-space methods; Terminology;
Conference_Titel :
Neural Networks, 2002. IJCNN '02. Proceedings of the 2002 International Joint Conference on
Conference_Location :
Honolulu, HI
Print_ISBN :
0-7803-7278-6
DOI :
10.1109/IJCNN.2002.1005433
