DocumentCode
298999
Title
Convergence properties of self-organizing neural networks
Author
Horowitz, Roberto ; Alvarez, Luis
Author_Institution
Dept. of Mech. Eng., California Univ., Berkeley, CA, USA
Volume
2
fYear
1995
fDate
21-23 Jun 1995
Firstpage
1339
Abstract
In this paper we analyze the convergence properties of a class of self-organizing neural networks, introduced and popularized by Kohonen, using the ODE approach. It is shown that Kohonen´s learning law converges to the best locally affine feature map. A new integrally distributed self-organizing learning law is presented which converges to the equiprobable feature map for inputs which have arbitrary random probability distribution functions
Keywords
differential equations; learning (artificial intelligence); probability; self-organising feature maps; Kohonen neural nets; ODE approach; best locally affine feature map; convergence properties; equiprobable feature map; integrally distributed self-organizing learning law; ordinary differential equations; self-organizing neural networks; Adaptive systems; Convergence; Electronic mail; Input variables; Markov processes; Mechanical engineering; Mechanical factors; Network topology; Neural networks; Probability distribution;
fLanguage
English
Publisher
ieee
Conference_Titel
American Control Conference, Proceedings of the 1995
Conference_Location
Seattle, WA
Print_ISBN
0-7803-2445-5
Type
conf
DOI
10.1109/ACC.1995.520968
Filename
520968
Link To Document