Evaluation of neural network robust reliability using information-gap theory

S.G. Pierce, Y. Ben-Haim, K. Worden, G. Manson

Research output: Contribution to journalArticle

28 Citations (Scopus)

Abstract

A novel technique for the evaluation of neural network robustness against uncertainty using a nonprobabilistic approach is presented. Conventional optimization techniques were employed to train multilayer perceptron (MLP) networks, which were then probed with an uncertainty analysis using an information-gap model to quantify the network response to uncertainty in
the input data. It is demonstrated that the best performing network on data with low uncertainty is not in general the optimal network on data with a higher degree of input uncertainty. Using the concepts of information-gap theory, this paper develops a theoretical framework for information-gap uncertainty applied to neural networks, and explores the practical application of the procedure to
three sample cases. The first consists of a simple two-dimensional (2-D) classification network operating on a known Gaussian distribution, the second a nine-lass vibration classification problem from an aircraft wing, and the third a two-class example from a database of breast cancer incidence.
LanguageEnglish
Pages1349-1361
Number of pages12
JournalIEEE Transactions on Neural Networks
Volume17
Issue number6
DOIs
Publication statusPublished - 2006

Fingerprint

Neural networks
Uncertainty analysis
Gaussian distribution
Multilayer neural networks
Uncertainty

Keywords

  • Information-gap
  • robustness
  • neural networks
  • uncertainty
  • gaussian distribution
  • vibration classification
  • uncertainty analysis
  • nonprobabilistic method
  • neural network robust reliability
  • multilayer perceptron network
  • information-gap theory
  • breast cancer incidence
  • aircraft wing

Cite this

Pierce, S.G. ; Ben-Haim, Y. ; Worden, K. ; Manson, G. / Evaluation of neural network robust reliability using information-gap theory. In: IEEE Transactions on Neural Networks. 2006 ; Vol. 17, No. 6. pp. 1349-1361.
@article{df7d818c91e44387b5ab8f26b1034590,
title = "Evaluation of neural network robust reliability using information-gap theory",
abstract = "A novel technique for the evaluation of neural network robustness against uncertainty using a nonprobabilistic approach is presented. Conventional optimization techniques were employed to train multilayer perceptron (MLP) networks, which were then probed with an uncertainty analysis using an information-gap model to quantify the network response to uncertainty inthe input data. It is demonstrated that the best performing network on data with low uncertainty is not in general the optimal network on data with a higher degree of input uncertainty. Using the concepts of information-gap theory, this paper develops a theoretical framework for information-gap uncertainty applied to neural networks, and explores the practical application of the procedure tothree sample cases. The first consists of a simple two-dimensional (2-D) classification network operating on a known Gaussian distribution, the second a nine-lass vibration classification problem from an aircraft wing, and the third a two-class example from a database of breast cancer incidence.",
keywords = "Information-gap, robustness, neural networks, uncertainty , gaussian distribution, vibration classification, uncertainty analysis, nonprobabilistic method, neural network robust reliability, multilayer perceptron network, information-gap theory, breast cancer incidence, aircraft wing",
author = "S.G. Pierce and Y. Ben-Haim and K. Worden and G. Manson",
year = "2006",
doi = "10.1109/TNN.2006.880363",
language = "English",
volume = "17",
pages = "1349--1361",
journal = "IEEE Transactions on Neural Networks",
issn = "1045-9227",
number = "6",

}

Evaluation of neural network robust reliability using information-gap theory. / Pierce, S.G.; Ben-Haim, Y.; Worden, K.; Manson, G.

In: IEEE Transactions on Neural Networks, Vol. 17, No. 6, 2006, p. 1349-1361.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Evaluation of neural network robust reliability using information-gap theory

AU - Pierce, S.G.

AU - Ben-Haim, Y.

AU - Worden, K.

AU - Manson, G.

PY - 2006

Y1 - 2006

N2 - A novel technique for the evaluation of neural network robustness against uncertainty using a nonprobabilistic approach is presented. Conventional optimization techniques were employed to train multilayer perceptron (MLP) networks, which were then probed with an uncertainty analysis using an information-gap model to quantify the network response to uncertainty inthe input data. It is demonstrated that the best performing network on data with low uncertainty is not in general the optimal network on data with a higher degree of input uncertainty. Using the concepts of information-gap theory, this paper develops a theoretical framework for information-gap uncertainty applied to neural networks, and explores the practical application of the procedure tothree sample cases. The first consists of a simple two-dimensional (2-D) classification network operating on a known Gaussian distribution, the second a nine-lass vibration classification problem from an aircraft wing, and the third a two-class example from a database of breast cancer incidence.

AB - A novel technique for the evaluation of neural network robustness against uncertainty using a nonprobabilistic approach is presented. Conventional optimization techniques were employed to train multilayer perceptron (MLP) networks, which were then probed with an uncertainty analysis using an information-gap model to quantify the network response to uncertainty inthe input data. It is demonstrated that the best performing network on data with low uncertainty is not in general the optimal network on data with a higher degree of input uncertainty. Using the concepts of information-gap theory, this paper develops a theoretical framework for information-gap uncertainty applied to neural networks, and explores the practical application of the procedure tothree sample cases. The first consists of a simple two-dimensional (2-D) classification network operating on a known Gaussian distribution, the second a nine-lass vibration classification problem from an aircraft wing, and the third a two-class example from a database of breast cancer incidence.

KW - Information-gap

KW - robustness

KW - neural networks

KW - uncertainty

KW - gaussian distribution

KW - vibration classification

KW - uncertainty analysis

KW - nonprobabilistic method

KW - neural network robust reliability

KW - multilayer perceptron network

KW - information-gap theory

KW - breast cancer incidence

KW - aircraft wing

UR - http://dx.doi.org/10.1109/TNN.2006.880363

U2 - 10.1109/TNN.2006.880363

DO - 10.1109/TNN.2006.880363

M3 - Article

VL - 17

SP - 1349

EP - 1361

JO - IEEE Transactions on Neural Networks

T2 - IEEE Transactions on Neural Networks

JF - IEEE Transactions on Neural Networks

SN - 1045-9227

IS - 6

ER -