Efficient training of RBF networks for classification.

Ian T. Nabney

Research output: Contribution to journalArticle

Abstract

Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. We show how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm. This approach can be used with any model which consists of a generalised linear output function applied to a model which is linear in its parameters. We compare this approach with standard non-linear optimisation algorithms on a number of datasets.

Original languageEnglish
Pages (from-to)201-208
Number of pages8
JournalInternational Journal of Neural Systems
Volume14
Issue number3
DOIs
Publication statusPublished - Jun 2004

Fingerprint

Radial basis function networks
Multilayer neural networks
Logistics

Bibliographical note

Electronic version of an article published as International Journal of Neural Systems, 14 (3), 2004, pp. 201-208, Article DOI: 10.1142/S0129065704001930 © World Scientific Publishing Company http://www.worldscinet.com/ijns/ijns.shtml

Keywords

  • radial basis function
  • non-linear optimisation
  • probabilistic modelling
  • classification

Cite this

@article{bccd781ed843404aa3869b874e67ba24,
title = "Efficient training of RBF networks for classification.",
abstract = "Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. We show how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm. This approach can be used with any model which consists of a generalised linear output function applied to a model which is linear in its parameters. We compare this approach with standard non-linear optimisation algorithms on a number of datasets.",
keywords = "radial basis function, non-linear optimisation, probabilistic modelling, classification",
author = "Nabney, {Ian T.}",
note = "Electronic version of an article published as International Journal of Neural Systems, 14 (3), 2004, pp. 201-208, Article DOI: 10.1142/S0129065704001930 {\circledC} World Scientific Publishing Company http://www.worldscinet.com/ijns/ijns.shtml",
year = "2004",
month = "6",
doi = "10.1142/S0129065704001930",
language = "English",
volume = "14",
pages = "201--208",
journal = "International Journal of Neural Systems",
issn = "0129-0657",
publisher = "World Scientific",
number = "3",

}

Efficient training of RBF networks for classification. / Nabney, Ian T.

In: International Journal of Neural Systems, Vol. 14, No. 3, 06.2004, p. 201-208.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Efficient training of RBF networks for classification.

AU - Nabney, Ian T.

N1 - Electronic version of an article published as International Journal of Neural Systems, 14 (3), 2004, pp. 201-208, Article DOI: 10.1142/S0129065704001930 © World Scientific Publishing Company http://www.worldscinet.com/ijns/ijns.shtml

PY - 2004/6

Y1 - 2004/6

N2 - Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. We show how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm. This approach can be used with any model which consists of a generalised linear output function applied to a model which is linear in its parameters. We compare this approach with standard non-linear optimisation algorithms on a number of datasets.

AB - Radial Basis Function networks with linear outputs are often used in regression problems because they can be substantially faster to train than Multi-layer Perceptrons. For classification problems, the use of linear outputs is less appropriate as the outputs are not guaranteed to represent probabilities. We show how RBFs with logistic and softmax outputs can be trained efficiently using the Fisher scoring algorithm. This approach can be used with any model which consists of a generalised linear output function applied to a model which is linear in its parameters. We compare this approach with standard non-linear optimisation algorithms on a number of datasets.

KW - radial basis function

KW - non-linear optimisation

KW - probabilistic modelling

KW - classification

UR - http://www.scopus.com/inward/record.url?scp=16544373696&partnerID=8YFLogxK

UR - https://www.worldscientific.com/doi/abs/10.1142/S0129065704001930

U2 - 10.1142/S0129065704001930

DO - 10.1142/S0129065704001930

M3 - Article

VL - 14

SP - 201

EP - 208

JO - International Journal of Neural Systems

JF - International Journal of Neural Systems

SN - 0129-0657

IS - 3

ER -