The University of Southampton
University of Southampton Institutional Repository

Neurofuzzy Model Weight Identification With Multiple Priors

Neurofuzzy Model Weight Identification With Multiple Priors
Neurofuzzy Model Weight Identification With Multiple Priors
The structure of neurofuzzy systems is restricted by the need for a fuzzy rule interpretation. This often results in some redundant structure, which is hard to identify using conventional ML estimation i.e. conventional supervised learning. This paper investigates the application of regularisation techniques to these neurofuzzy models to help improve their generalisation capabilities. In particular, this theory is applied to the additive neurofuzzy structure identified by B-spline neurofuzzy construction algorithms. Bayesian inferencing techniques in the form of MAP estimation are applied to these models resulting regularisation and an effective method for identifying the regularisation coefficient (or hyperparameters) i.e. evidence maximisation is derived. These techniques are extended to local regularisation, where a weight prior is defined for each submodel. The construction of these priors in both global and local regularisation is described. Two methods are proposed for the identification of the multiple hyperparameters: evidence maximisation and a method combining backfitting and conventional evidence maximisation techniques. These are both shown to work well on a numerical example, but due to seemly correlated inputs backfitting takes significantly longer to converge.
Bossley, K.M.
de1a2979-b9e9-481e-af09-0b4887f0f360
Brown, M.
52cf4f52-6839-4658-8cc5-ec51da626049
Harris, C.J.
c4fd3763-7b3f-4db1-9ca3-5501080f797a
Bossley, K.M.
de1a2979-b9e9-481e-af09-0b4887f0f360
Brown, M.
52cf4f52-6839-4658-8cc5-ec51da626049
Harris, C.J.
c4fd3763-7b3f-4db1-9ca3-5501080f797a

Bossley, K.M., Brown, M. and Harris, C.J. (1996) Neurofuzzy Model Weight Identification With Multiple Priors. IEEE Trans Neural Networks.

Record type: Article

Abstract

The structure of neurofuzzy systems is restricted by the need for a fuzzy rule interpretation. This often results in some redundant structure, which is hard to identify using conventional ML estimation i.e. conventional supervised learning. This paper investigates the application of regularisation techniques to these neurofuzzy models to help improve their generalisation capabilities. In particular, this theory is applied to the additive neurofuzzy structure identified by B-spline neurofuzzy construction algorithms. Bayesian inferencing techniques in the form of MAP estimation are applied to these models resulting regularisation and an effective method for identifying the regularisation coefficient (or hyperparameters) i.e. evidence maximisation is derived. These techniques are extended to local regularisation, where a weight prior is defined for each submodel. The construction of these priors in both global and local regularisation is described. Two methods are proposed for the identification of the multiple hyperparameters: evidence maximisation and a method combining backfitting and conventional evidence maximisation techniques. These are both shown to work well on a numerical example, but due to seemly correlated inputs backfitting takes significantly longer to converge.

Full text not available from this repository.

More information

Published date: 1996
Additional Information: submitted for publication
Organisations: Southampton Wireless Group

Identifiers

Local EPrints ID: 250117
URI: https://eprints.soton.ac.uk/id/eprint/250117
PURE UUID: fad4b848-9ea3-4f24-98fd-4b27f8acc024

Catalogue record

Date deposited: 04 May 1999
Last modified: 18 Jul 2017 10:44

Export record

Download statistics

Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.

View more statistics

Atom RSS 1.0 RSS 2.0

Contact ePrints Soton: eprints@soton.ac.uk

ePrints Soton supports OAI 2.0 with a base URL of https://eprints.soton.ac.uk/cgi/oai2

This repository has been built using EPrints software, developed at the University of Southampton, but available to everyone to use.

We use cookies to ensure that we give you the best experience on our website. If you continue without changing your settings, we will assume that you are happy to receive cookies on the University of Southampton website.

×