An online learning network for biometric scores fusion

Youngsung Kim, Kar Ann Toh, Beng Jin Teoh, How Lung Eng, Wei Yun Yau

Research output: Contribution to journalArticle

16 Citations (Scopus)

Abstract

In design of a multibiometric system, a major concern is the learning cost in terms of computation complexity and memory usage due to large data size. In this paper, we propose an online learning network to circumvent the computational problem. Although conventional online learning algorithms can be adopted, their optimization of the fitting distance residuals does not meet the actual classification error requirement. A direct optimization to the classification performance is thus desired. Since the proposed classification-based formulation involves a class-specific weight which varies according to the total number of genuine-users and imposters, an online learning formulation becomes non-trivial. Extensive empirical evaluations on publicly available data sets show promising potential of the proposed method in terms of fusion verification accuracy and computational cost.

Original languageEnglish
Pages (from-to)65-77
Number of pages13
JournalNeurocomputing
Volume102
DOIs
Publication statusPublished - 2013 Feb 15

Fingerprint

Biometrics
Fusion reactions
Learning
Costs and Cost Analysis
Density (specific gravity)
Learning algorithms
Costs
Data storage equipment
Weights and Measures

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

Kim, Youngsung ; Toh, Kar Ann ; Teoh, Beng Jin ; Eng, How Lung ; Yau, Wei Yun. / An online learning network for biometric scores fusion. In: Neurocomputing. 2013 ; Vol. 102. pp. 65-77.
@article{0e79c171644e4171bdcdf6fc7398fc04,
title = "An online learning network for biometric scores fusion",
abstract = "In design of a multibiometric system, a major concern is the learning cost in terms of computation complexity and memory usage due to large data size. In this paper, we propose an online learning network to circumvent the computational problem. Although conventional online learning algorithms can be adopted, their optimization of the fitting distance residuals does not meet the actual classification error requirement. A direct optimization to the classification performance is thus desired. Since the proposed classification-based formulation involves a class-specific weight which varies according to the total number of genuine-users and imposters, an online learning formulation becomes non-trivial. Extensive empirical evaluations on publicly available data sets show promising potential of the proposed method in terms of fusion verification accuracy and computational cost.",
author = "Youngsung Kim and Toh, {Kar Ann} and Teoh, {Beng Jin} and Eng, {How Lung} and Yau, {Wei Yun}",
year = "2013",
month = "2",
day = "15",
doi = "10.1016/j.neucom.2011.12.048",
language = "English",
volume = "102",
pages = "65--77",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",

}

An online learning network for biometric scores fusion. / Kim, Youngsung; Toh, Kar Ann; Teoh, Beng Jin; Eng, How Lung; Yau, Wei Yun.

In: Neurocomputing, Vol. 102, 15.02.2013, p. 65-77.

Research output: Contribution to journalArticle

TY - JOUR

T1 - An online learning network for biometric scores fusion

AU - Kim, Youngsung

AU - Toh, Kar Ann

AU - Teoh, Beng Jin

AU - Eng, How Lung

AU - Yau, Wei Yun

PY - 2013/2/15

Y1 - 2013/2/15

N2 - In design of a multibiometric system, a major concern is the learning cost in terms of computation complexity and memory usage due to large data size. In this paper, we propose an online learning network to circumvent the computational problem. Although conventional online learning algorithms can be adopted, their optimization of the fitting distance residuals does not meet the actual classification error requirement. A direct optimization to the classification performance is thus desired. Since the proposed classification-based formulation involves a class-specific weight which varies according to the total number of genuine-users and imposters, an online learning formulation becomes non-trivial. Extensive empirical evaluations on publicly available data sets show promising potential of the proposed method in terms of fusion verification accuracy and computational cost.

AB - In design of a multibiometric system, a major concern is the learning cost in terms of computation complexity and memory usage due to large data size. In this paper, we propose an online learning network to circumvent the computational problem. Although conventional online learning algorithms can be adopted, their optimization of the fitting distance residuals does not meet the actual classification error requirement. A direct optimization to the classification performance is thus desired. Since the proposed classification-based formulation involves a class-specific weight which varies according to the total number of genuine-users and imposters, an online learning formulation becomes non-trivial. Extensive empirical evaluations on publicly available data sets show promising potential of the proposed method in terms of fusion verification accuracy and computational cost.

UR - http://www.scopus.com/inward/record.url?scp=84870242938&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84870242938&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2011.12.048

DO - 10.1016/j.neucom.2011.12.048

M3 - Article

VL - 102

SP - 65

EP - 77

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

ER -