Title :
Properties of a chaotic network separating memory patterns
Author :
Matykiewicz, Pawel
Author_Institution :
Dept. of Inf., Nicholas Copernicus Univ., Torun, Poland
Abstract :
A simple method aimed at improving the separation abilities of a chaotic neural network is presented and its memory properties investigated. Estimation of the invulnerability to the external input disturbance and the damage of weight connections are performed. Significant improvements of retrieval characteristic are reported. When weight connections are damaged, high instability of separation of the memory patterns is observed.
Keywords :
chaos; content-addressable storage; neural nets; pattern classification; chaotic neural network; content-addressable storage; external input disturbance; memory pattern separation; weight connection damages; Associative memory; Chaos; Distortion measurement; Electronic mail; Equations; Informatics; Information representation; Neural networks; Neurons; Object recognition;
Conference_Titel :
Neural Networks, 2004. Proceedings. 2004 IEEE International Joint Conference on
Print_ISBN :
0-7803-8359-1
DOI :
10.1109/IJCNN.2004.1380055
