DocumentCode :
1626071
Title :
New entropy learning method for neural network
Author :
Chan, Khue Hiang ; Ng, Geok See ; Erdogan, Sevki S. ; Singh, Harcharan
Author_Institution :
Sch. of Appl. Sci., Nanyang Technol. Univ., Singapore
Volume :
3
fYear :
1999
fDate :
6/21/1905 12:00:00 AM
Firstpage :
412
Abstract :
An entropy penalty term is used to steer the direction of the hidden node´s activation in the process of learning. A state with minimum entropy means that nodes are operating near the extreme values of the Sigmoid curve. As the training proceeds, redundant hidden nodes´ activations are pushed towards their extreme value, while relevant nodes remain active in the linear region of the Sigmoid curve. The early creation of redundant nodes may impair generalisation. To prevent the network from being driven into saturation before it can really learn, an entropy cycle is proposed to dampen the early creation of such redundant nodes
Keywords :
entropy; generalisation (artificial intelligence); learning (artificial intelligence); neural nets; Sigmoid curve; entropy cycle; entropy learning method; hidden node activation; linear region; minimum entropy; redundant node activations; saturation; Computational efficiency; Cost function; Differential equations; Entropy; Learning systems; Neural networks;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Systems, Man, and Cybernetics, 1999. IEEE SMC '99 Conference Proceedings. 1999 IEEE International Conference on
Conference_Location :
Tokyo
ISSN :
1062-922X
Print_ISBN :
0-7803-5731-0
Type :
conf
DOI :
10.1109/ICSMC.1999.823240
Filename :
823240
Link To Document :
بازگشت