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 paperpeer-review

    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
    DOIs
    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
    Country/TerritoryChina
    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