Title :
A new analytical method for AM neural networks and its application to pattern recognition
Author :
Zhenjiang, Mia0 ; Baozong, Yuan
Author_Institution :
Inst. of Inf. Sci., Northern Jiatong Univ., Beijing, China
Abstract :
The problems of asymptotical stability and associative memory (AM) for Hopfield-type neural networks are analyzed. A new kind of energy function is introduced which is different from Hopfield´s definition. The networks´ asymptotical stability is analyzed by means of the new energy function. Two theorems are obtained. By comparing these two theorems with the existing conclusions it can be found that, in some cases, they are consistent and, in other cases, they are not equivalent but rather complement each other. By using these theorems, one associative memory neural network is designed, and their applications to pattern recognition are demonstrated
Keywords :
Hopfield neural nets; content-addressable storage; image recognition; Hopfield-type neural networks; associative memory neural nets; asymptotical stability; energy function; pattern recognition; Artificial neural networks; Associative memory; Asymptotic stability; Hopfield neural networks; Information analysis; Jacobian matrices; Neural networks; Pattern analysis; Pattern recognition; Stability analysis;
Conference_Titel :
Neural Networks, 1993., IEEE International Conference on
Conference_Location :
San Francisco, CA
Print_ISBN :
0-7803-0999-5
DOI :
10.1109/ICNN.1993.298790