On functional approximation with normalized Gaussian units

Authors:
Michel Benaim
Affiliations:
-
Venue:
Neural Computation
Year:
1994

Citing 0
Cited 3

Analysis of the Functional Block Involved in the Design of Radial Basis Function Networks

Neural Processing Letters
Nonlinear prediction of chaotic signals using a normalised radial basis function network

Signal Processing
Natural discriminant analysis using interactive Potts models

Neural Computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

Feedforward neural networks with a single hidden layer usingnormalized gaussian units are studied. It is proved that suchneural networks are capable of universal approximation in asatisfactory sense. Then, a hybrid learning rule as per Moody andDarken that combines unsupervised learning of hidden units andsupervised learning of output units is considered. By using themethod of ordinary differential equations for adaptive algorithms(ODE method) it is shown that the asymptotic properties of thelearning rule may be studied in terms of an autonomous cascade ofdynamical systems. Some recent results from Hirsch about cascadesare used to show the asymptotic stability of the learning rule.