(26) * << * >> * Russian * English * Content * All Issues

Stage-wise learning of radial neural networks

V.A. Fursov1,2, N.E. Kozin1,2
1Image Processing Systems Institute of RAS 

2Samara State Aerospace University 

 PDF, 10 kB

Pages: 138-141.

Abstract:
A two-stage procedure for the training of radial neural networks in pattern recognition is considered. At the first stage, the set of training samples is split into subsets and the class centers are adjusted for them. Then the parameters of the radial functions are adjusted for each subset of the training samples. The coefficients of contingency of the vectors of attributes of training samples are used to form subsets.

Keywords:
neural network, pattern recognition, radial function, vector of attribute.

Citation:
Fursov VA, Kozin NE. Stage-wise learning of radial neural networks. Computer Optics 2004; 26: 138-141.

References:
  1. Osowski S. Sieci neuronowe do przetwarzania informacji. Warszawa: Oficyna Wydawnicza Politechniki Warszawskiej; 2000. 
  2. Soifer VA, ed. Methods for computer design of diffractive optical elements. New York: John Willey & Sons Inc; 2002. 
  3. Fursov VA. Constructing unified identification algorithms using a small number of observations for adaptive control and navigation systems. Proc SPIE 1997; 3087: 34-44.

© 2009, IPSI RAS
151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail: ko@smr.ru ; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20