New radial basis function neural network training for nonlinear and nonstationary signals

Seng Kah Phooi, L. M. Ang

Research output: Book chapter/Published conference paperConference paper

Abstract

This paper deals with the problem of adaptation of radial basis function neural networks (RBF NN). A new RBF NN supervised training algorithm is proposed. This method possesses the distinctive properties of Lyapunov Theory-based Adaptive Filtering (LAF) in [1]-[2]. The method is different from many RBF NN training using gradient search methods. A new Lyapunov function of the error between the desired output and the RBF NN output is first defined. The output asymptotically converges to the desired output by designing the adaptation law in Lyapunov sense. Error convergence analysis in this paper has proven that the design of the new RBF NN training algorithm is independent of statistic properties of input and output signals. The new adaptation law has better tracking capability compared with the tracking performance of LAF in [1]-[2]. The performance of the proposed technique is illustrated through the adaptive prediction of nonlinear and nonstationary speech signals.

Original languageEnglish
Title of host publicationComputational Intelligence and Security - International Conference, CIS 2006, Revised Selected Papers
Pages220-230
Number of pages11
Publication statusPublished - 01 Dec 2007
EventInternational Conference on Computational Intelligence and Security, CIS 2006 - Guangzhou, China
Duration: 03 Nov 200606 Nov 2006

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume4456 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceInternational Conference on Computational Intelligence and Security, CIS 2006
CountryChina
CityGuangzhou
Period03/11/0606/11/06

Fingerprint Dive into the research topics of 'New radial basis function neural network training for nonlinear and nonstationary signals'. Together they form a unique fingerprint.

Cite this