Learning from target knowledge approximation

Research output: Chapter in Book/Report/Conference proceedingConference contribution

15 Citations (Scopus)

Abstract

The classical Least Squares Error (LSE) learning method for pattern classification is to learn a classifier based on data density where the learning process (density-fitting error minimization) and the learning objective (classification error rate) do not find a good match. In this work, we propose to learn according to classification decision objectives directly. We shall work on two classification objectives namely, the Total Error Rate and the Receiver Operating Characteristics, and directly optimize the learning process according to these objectives. Using a learning model which is linear in its parameters, we propose two approximation methods to optimize these classification objectives. Our empirical results on biometrics fusion show comparable performances of the proposed methods with the widely used Support Vector Machines (SVM), with one of the approaches having a clear advantage of fast single-step solution.

Original languageEnglish
Title of host publication2006 1st IEEE Conference on Industrial Electronics and Applications
DOIs
Publication statusPublished - 2006 Dec 1
Event2006 1st IEEE Conference on Industrial Electronics and Applications, ICIEA 2006 - Singapore, Singapore
Duration: 2006 May 242006 May 26

Other

Other2006 1st IEEE Conference on Industrial Electronics and Applications, ICIEA 2006
CountrySingapore
CitySingapore
Period06/5/2406/5/26

Fingerprint

Biometrics
Pattern recognition
Support vector machines
Classifiers
Fusion reactions

All Science Journal Classification (ASJC) codes

  • Electrical and Electronic Engineering
  • Industrial and Manufacturing Engineering

Cite this

Toh, K. A. (2006). Learning from target knowledge approximation. In 2006 1st IEEE Conference on Industrial Electronics and Applications [4026003] https://doi.org/10.1109/ICIEA.2006.257074
Toh, Kar Ann. / Learning from target knowledge approximation. 2006 1st IEEE Conference on Industrial Electronics and Applications. 2006.
@inproceedings{32b20f5f05c340bd92f09e8e07194c8f,
title = "Learning from target knowledge approximation",
abstract = "The classical Least Squares Error (LSE) learning method for pattern classification is to learn a classifier based on data density where the learning process (density-fitting error minimization) and the learning objective (classification error rate) do not find a good match. In this work, we propose to learn according to classification decision objectives directly. We shall work on two classification objectives namely, the Total Error Rate and the Receiver Operating Characteristics, and directly optimize the learning process according to these objectives. Using a learning model which is linear in its parameters, we propose two approximation methods to optimize these classification objectives. Our empirical results on biometrics fusion show comparable performances of the proposed methods with the widely used Support Vector Machines (SVM), with one of the approaches having a clear advantage of fast single-step solution.",
author = "Toh, {Kar Ann}",
year = "2006",
month = "12",
day = "1",
doi = "10.1109/ICIEA.2006.257074",
language = "English",
isbn = "078039514X",
booktitle = "2006 1st IEEE Conference on Industrial Electronics and Applications",

}

Toh, KA 2006, Learning from target knowledge approximation. in 2006 1st IEEE Conference on Industrial Electronics and Applications., 4026003, 2006 1st IEEE Conference on Industrial Electronics and Applications, ICIEA 2006, Singapore, Singapore, 06/5/24. https://doi.org/10.1109/ICIEA.2006.257074

Learning from target knowledge approximation. / Toh, Kar Ann.

2006 1st IEEE Conference on Industrial Electronics and Applications. 2006. 4026003.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Learning from target knowledge approximation

AU - Toh, Kar Ann

PY - 2006/12/1

Y1 - 2006/12/1

N2 - The classical Least Squares Error (LSE) learning method for pattern classification is to learn a classifier based on data density where the learning process (density-fitting error minimization) and the learning objective (classification error rate) do not find a good match. In this work, we propose to learn according to classification decision objectives directly. We shall work on two classification objectives namely, the Total Error Rate and the Receiver Operating Characteristics, and directly optimize the learning process according to these objectives. Using a learning model which is linear in its parameters, we propose two approximation methods to optimize these classification objectives. Our empirical results on biometrics fusion show comparable performances of the proposed methods with the widely used Support Vector Machines (SVM), with one of the approaches having a clear advantage of fast single-step solution.

AB - The classical Least Squares Error (LSE) learning method for pattern classification is to learn a classifier based on data density where the learning process (density-fitting error minimization) and the learning objective (classification error rate) do not find a good match. In this work, we propose to learn according to classification decision objectives directly. We shall work on two classification objectives namely, the Total Error Rate and the Receiver Operating Characteristics, and directly optimize the learning process according to these objectives. Using a learning model which is linear in its parameters, we propose two approximation methods to optimize these classification objectives. Our empirical results on biometrics fusion show comparable performances of the proposed methods with the widely used Support Vector Machines (SVM), with one of the approaches having a clear advantage of fast single-step solution.

UR - http://www.scopus.com/inward/record.url?scp=42749098590&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=42749098590&partnerID=8YFLogxK

U2 - 10.1109/ICIEA.2006.257074

DO - 10.1109/ICIEA.2006.257074

M3 - Conference contribution

SN - 078039514X

SN - 9780780395145

BT - 2006 1st IEEE Conference on Industrial Electronics and Applications

ER -

Toh KA. Learning from target knowledge approximation. In 2006 1st IEEE Conference on Industrial Electronics and Applications. 2006. 4026003 https://doi.org/10.1109/ICIEA.2006.257074