Classification using linear models with uncertain weights

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

Research output: Contribution to conferencePaper

Abstract

Linear models were trained on a simple two-class two-dimensional data set. The network connections, which were crisp, real numbers, were then replaced with interval ranges of real numbers. Crisp input data was propagated through these uncertain-weighted networks to give interval ranges on the output values. The classification rates of the networks could be adjusted by the level of uncertainty in the connections, allowing the user to specify an acceptable misclassification rate and choosing the network with the best corresponding correct classification rate. Vertex propagation, interval arithmetic and affine arithmetic were used to represent the uncertainty in networks with linear and softmax output activation functions and were benchmarked against a simple output threshold approach. It was found that, although the network responses varied, the various techniques returned similar relative classification rates.
LanguageEnglish
Pages969-974
Number of pages5
Publication statusPublished - Sep 2005
EventEurodyn 2005: 6th International Conference on Structural Dynamics - Paris, France
Duration: 4 Sep 20057 Sep 2005

Conference

ConferenceEurodyn 2005: 6th International Conference on Structural Dynamics
CountryFrance
CityParis
Period4/09/057/09/05

Fingerprint

rate

Keywords

  • classification
  • linear models
  • uncertain weights

Cite this

Manson, G., Pierce, S. G., Worden, K., & Chetwynd, D. (2005). Classification using linear models with uncertain weights. 969-974. Paper presented at Eurodyn 2005: 6th International Conference on Structural Dynamics, Paris, France.
Manson, G. ; Pierce, S.G. ; Worden, K. ; Chetwynd, D. / Classification using linear models with uncertain weights. Paper presented at Eurodyn 2005: 6th International Conference on Structural Dynamics, Paris, France.5 p.
@conference{dda7549b0fe84d2f8db95ef80e272684,
title = "Classification using linear models with uncertain weights",
abstract = "Linear models were trained on a simple two-class two-dimensional data set. The network connections, which were crisp, real numbers, were then replaced with interval ranges of real numbers. Crisp input data was propagated through these uncertain-weighted networks to give interval ranges on the output values. The classification rates of the networks could be adjusted by the level of uncertainty in the connections, allowing the user to specify an acceptable misclassification rate and choosing the network with the best corresponding correct classification rate. Vertex propagation, interval arithmetic and affine arithmetic were used to represent the uncertainty in networks with linear and softmax output activation functions and were benchmarked against a simple output threshold approach. It was found that, although the network responses varied, the various techniques returned similar relative classification rates.",
keywords = "classification, linear models, uncertain weights",
author = "G. Manson and S.G. Pierce and K. Worden and D. Chetwynd",
year = "2005",
month = "9",
language = "English",
pages = "969--974",
note = "Eurodyn 2005: 6th International Conference on Structural Dynamics ; Conference date: 04-09-2005 Through 07-09-2005",

}

Manson, G, Pierce, SG, Worden, K & Chetwynd, D 2005, 'Classification using linear models with uncertain weights' Paper presented at Eurodyn 2005: 6th International Conference on Structural Dynamics, Paris, France, 4/09/05 - 7/09/05, pp. 969-974.

Classification using linear models with uncertain weights. / Manson, G.; Pierce, S.G.; Worden, K.; Chetwynd, D.

2005. 969-974 Paper presented at Eurodyn 2005: 6th International Conference on Structural Dynamics, Paris, France.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Classification using linear models with uncertain weights

AU - Manson, G.

AU - Pierce, S.G.

AU - Worden, K.

AU - Chetwynd, D.

PY - 2005/9

Y1 - 2005/9

N2 - Linear models were trained on a simple two-class two-dimensional data set. The network connections, which were crisp, real numbers, were then replaced with interval ranges of real numbers. Crisp input data was propagated through these uncertain-weighted networks to give interval ranges on the output values. The classification rates of the networks could be adjusted by the level of uncertainty in the connections, allowing the user to specify an acceptable misclassification rate and choosing the network with the best corresponding correct classification rate. Vertex propagation, interval arithmetic and affine arithmetic were used to represent the uncertainty in networks with linear and softmax output activation functions and were benchmarked against a simple output threshold approach. It was found that, although the network responses varied, the various techniques returned similar relative classification rates.

AB - Linear models were trained on a simple two-class two-dimensional data set. The network connections, which were crisp, real numbers, were then replaced with interval ranges of real numbers. Crisp input data was propagated through these uncertain-weighted networks to give interval ranges on the output values. The classification rates of the networks could be adjusted by the level of uncertainty in the connections, allowing the user to specify an acceptable misclassification rate and choosing the network with the best corresponding correct classification rate. Vertex propagation, interval arithmetic and affine arithmetic were used to represent the uncertainty in networks with linear and softmax output activation functions and were benchmarked against a simple output threshold approach. It was found that, although the network responses varied, the various techniques returned similar relative classification rates.

KW - classification

KW - linear models

KW - uncertain weights

M3 - Paper

SP - 969

EP - 974

ER -

Manson G, Pierce SG, Worden K, Chetwynd D. Classification using linear models with uncertain weights. 2005. Paper presented at Eurodyn 2005: 6th International Conference on Structural Dynamics, Paris, France.