Evaluation of neural network performance and generalisation using thresholding functions

S.G. Pierce, K. Worden, G. Manson

Research output: Contribution to journalArticle

2 Citations (Scopus)

Abstract

The application of a simple thresholding technique to help assess the satisfactory performance of classification networks formed from Multi-Layer Perceptron (MLP) artificial neural networks (ANNs) is discussed. Both conventional Maximum Likelihood and Bayesian Evidence based training paradigms were
implemented. Firstly a simulated data set drawn from a two-dimensional Gaussian distribution was investigated to illustrate the physical significance of the threshold plots compared to the classifier output probability contours. Secondly a real world application data set comprising of low-frequency vibration measurements on an aircraft wing (a GNAT trainer) is considered. It is demonstrated that simple threshold based plots applied to classifier network outputs may provide a simple yet powerful technique to aid in the rejection of poorly regularised network structures.
LanguageEnglish
Pages109-124
Number of pages16
JournalNeural Computing and Applications
Volume16
Issue number2
Early online date31 May 2006
DOIs
Publication statusPublished - 2007

Fingerprint

Network performance
Classifiers
Neural networks
Vibration measurement
Gaussian distribution
Multilayer neural networks
Maximum likelihood

Keywords

  • neural network training
  • thresholding
  • classification networks

Cite this

@article{d778c4b78e154aec9d98fd4a439c298f,
title = "Evaluation of neural network performance and generalisation using thresholding functions",
abstract = "The application of a simple thresholding technique to help assess the satisfactory performance of classification networks formed from Multi-Layer Perceptron (MLP) artificial neural networks (ANNs) is discussed. Both conventional Maximum Likelihood and Bayesian Evidence based training paradigms were implemented. Firstly a simulated data set drawn from a two-dimensional Gaussian distribution was investigated to illustrate the physical significance of the threshold plots compared to the classifier output probability contours. Secondly a real world application data set comprising of low-frequency vibration measurements on an aircraft wing (a GNAT trainer) is considered. It is demonstrated that simple threshold based plots applied to classifier network outputs may provide a simple yet powerful technique to aid in the rejection of poorly regularised network structures.",
keywords = "neural network training, thresholding, classification networks",
author = "S.G. Pierce and K. Worden and G. Manson",
year = "2007",
doi = "10.1007/s00521-006-0044-z",
language = "English",
volume = "16",
pages = "109--124",
journal = "Neural Computing and Applications",
issn = "0941-0643",
number = "2",

}

Evaluation of neural network performance and generalisation using thresholding functions. / Pierce, S.G.; Worden, K.; Manson, G.

In: Neural Computing and Applications, Vol. 16, No. 2, 2007, p. 109-124.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Evaluation of neural network performance and generalisation using thresholding functions

AU - Pierce, S.G.

AU - Worden, K.

AU - Manson, G.

PY - 2007

Y1 - 2007

N2 - The application of a simple thresholding technique to help assess the satisfactory performance of classification networks formed from Multi-Layer Perceptron (MLP) artificial neural networks (ANNs) is discussed. Both conventional Maximum Likelihood and Bayesian Evidence based training paradigms were implemented. Firstly a simulated data set drawn from a two-dimensional Gaussian distribution was investigated to illustrate the physical significance of the threshold plots compared to the classifier output probability contours. Secondly a real world application data set comprising of low-frequency vibration measurements on an aircraft wing (a GNAT trainer) is considered. It is demonstrated that simple threshold based plots applied to classifier network outputs may provide a simple yet powerful technique to aid in the rejection of poorly regularised network structures.

AB - The application of a simple thresholding technique to help assess the satisfactory performance of classification networks formed from Multi-Layer Perceptron (MLP) artificial neural networks (ANNs) is discussed. Both conventional Maximum Likelihood and Bayesian Evidence based training paradigms were implemented. Firstly a simulated data set drawn from a two-dimensional Gaussian distribution was investigated to illustrate the physical significance of the threshold plots compared to the classifier output probability contours. Secondly a real world application data set comprising of low-frequency vibration measurements on an aircraft wing (a GNAT trainer) is considered. It is demonstrated that simple threshold based plots applied to classifier network outputs may provide a simple yet powerful technique to aid in the rejection of poorly regularised network structures.

KW - neural network training

KW - thresholding

KW - classification networks

U2 - 10.1007/s00521-006-0044-z

DO - 10.1007/s00521-006-0044-z

M3 - Article

VL - 16

SP - 109

EP - 124

JO - Neural Computing and Applications

T2 - Neural Computing and Applications

JF - Neural Computing and Applications

SN - 0941-0643

IS - 2

ER -