Self-organization sigmoidal blocks networks

Author

Valença, Mêuser ; Ludermir, Teresa

Author_Institution

Companhia Hidro-Eletrica do Sao Francisco, Brazil

Volume

3

fYear

1999

fDate

1999

Firstpage

1943

Abstract

We present a new class of higher-order feedforward neural networks, called self-organization sigmoidal blocks networks (SSBN). SSBN networks are based on the inductive learning method (also called self-organization). These new networks are shown to uniformly approximate any continuous function, with arbitrary degree of accuracy. The SSBN provides a natural mechanism for incremental network growth, and we develop a constructive algorithm based on the inductive learning method for the network. Simulation results of forecasting, approximation of nonlinear functions and approximation of multivariate polynomials are given, to highlight the capability of the network

Keywords

feedforward neural nets; forecasting theory; function approximation; learning by example; polynomial approximation; self-organising feature maps; feedforward neural networks; forecasting; higher-order neural networks; inductive learning; multivariate polynomials; nonlinear function approximation; self-organization sigmoidal blocks networks; Approximation algorithms; Data handling; Explosions; Feedforward neural networks; Function approximation; Learning systems; Mathematical model; Neural networks; Polynomials; Predictive models;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1999. IJCNN '99. International Joint Conference on

Conference_Location

Washington, DC

ISSN

1098-7576

Print_ISBN

0-7803-5529-6

Type

conf

DOI

10.1109/IJCNN.1999.832680

Filename

832680