Title :
A BP variant with improved convergence and generalization
Author :
Aloni-Lavi, Ruthi ; Metzger, Yaakov ; Stein, Yaakov
Author_Institution :
Efrat Future Technology Ltd., Tel Aviv, Israel
Abstract :
A variant of backpropagation that trains faster, gets caught in local minima less frequently, generalizes better for a given amount of training, and is specifically adapted to pattern recognition problems is presented. These abilities are demonstrated by simulations of an archetypal classification task
Keywords :
backpropagation; feedforward neural nets; pattern recognition; backpropagation; classification task; convergence; generalization; local minima; pattern recognition; Application software; Computer architecture; Convergence; Cost function; Dynamic programming; Feedforward neural networks; Nearest neighbor searches; Neural networks; Neurons; Pattern recognition;
Conference_Titel :
Neural Networks, 1992. IJCNN., International Joint Conference on
Conference_Location :
Baltimore, MD
Print_ISBN :
0-7803-0559-0
DOI :
10.1109/IJCNN.1992.287067
