Title :
A learning method of multilayer neural networks for pruning hidden units
Author :
Takechi, Hirochika ; Murakami, Kenji
Author_Institution :
Electrotech. Lab., Shikoku Res. Inst. Inc., Matsuyama, Japan
Abstract :
This paper presents a learning method of multilayer neural networks for pruning hidden units. In this method, the learning rates of each unit are adaptively changed to unbalance the influence of the hidden units on the network. With this method, useless hidden units are easily found and networks with small number of hidden units are constructed
Keywords :
adaptive systems; backpropagation; feedforward neural nets; backpropagation; hidden unit pruning; learning method; multilayer neural networks; Argon; Equations; Fault tolerance; Laboratories; Learning systems; Mean square error methods; Multi-layer neural network; Neural networks; Weight control;
Conference_Titel :
SICE '95. Proceedings of the 34th SICE Annual Conference. International Session Papers
Conference_Location :
Hokkaido
Print_ISBN :
0-7803-2781-0
DOI :
10.1109/SICE.1995.526961
