A sequential learning algorithm for self-adaptive resource allocation network classifier

S. Suresh, Keming Dong, Hyong Joong Kim

Research output: Contribution to journalArticle

99 Citations (Scopus)

Abstract

This paper addresses sequential learning algorithm for self-adaptive resource allocation network classifier. Our approach makes use of self-adaptive error based control parameters to alter the training data sequence, evolve the network architecture, and learn the network parameters. In addition, the algorithm removes the training samples which are similar to the stored knowledge in the network. Thereby, it avoids the over-training problem and reduces the training time significantly. Use of misclassification information and hinge loss error in growing/learning criterion helps in approximating the decision function accurately. The performance evaluation using balanced and imbalanced data sets shows that the proposed algorithm generates minimal network with lesser computation time to achieve higher classification performance.

Original languageEnglish
Pages (from-to)3012-3019
Number of pages8
JournalNeurocomputing
Volume73
Issue number16-18
DOIs
Publication statusPublished - 2010 Oct 1

Fingerprint

Resource Allocation
Learning algorithms
Resource allocation
Classifiers
Learning
Hinges
Network architecture

Keywords

  • Extended Kalman filter
  • Multi-category classification
  • Resource allocation network
  • Self-adaptive control parameters
  • Sequential learning

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Cognitive Neuroscience

Cite this

A sequential learning algorithm for self-adaptive resource allocation network classifier. / Suresh, S.; Dong, Keming; Kim, Hyong Joong.

In: Neurocomputing, Vol. 73, No. 16-18, 01.10.2010, p. 3012-3019.

Research output: Contribution to journalArticle

@article{cf77ec53c0084b2685842b3be7f8f3e2,
title = "A sequential learning algorithm for self-adaptive resource allocation network classifier",
abstract = "This paper addresses sequential learning algorithm for self-adaptive resource allocation network classifier. Our approach makes use of self-adaptive error based control parameters to alter the training data sequence, evolve the network architecture, and learn the network parameters. In addition, the algorithm removes the training samples which are similar to the stored knowledge in the network. Thereby, it avoids the over-training problem and reduces the training time significantly. Use of misclassification information and hinge loss error in growing/learning criterion helps in approximating the decision function accurately. The performance evaluation using balanced and imbalanced data sets shows that the proposed algorithm generates minimal network with lesser computation time to achieve higher classification performance.",
keywords = "Extended Kalman filter, Multi-category classification, Resource allocation network, Self-adaptive control parameters, Sequential learning",
author = "S. Suresh and Keming Dong and Kim, {Hyong Joong}",
year = "2010",
month = "10",
day = "1",
doi = "10.1016/j.neucom.2010.07.003",
language = "English",
volume = "73",
pages = "3012--3019",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "16-18",

}

TY - JOUR

T1 - A sequential learning algorithm for self-adaptive resource allocation network classifier

AU - Suresh, S.

AU - Dong, Keming

AU - Kim, Hyong Joong

PY - 2010/10/1

Y1 - 2010/10/1

N2 - This paper addresses sequential learning algorithm for self-adaptive resource allocation network classifier. Our approach makes use of self-adaptive error based control parameters to alter the training data sequence, evolve the network architecture, and learn the network parameters. In addition, the algorithm removes the training samples which are similar to the stored knowledge in the network. Thereby, it avoids the over-training problem and reduces the training time significantly. Use of misclassification information and hinge loss error in growing/learning criterion helps in approximating the decision function accurately. The performance evaluation using balanced and imbalanced data sets shows that the proposed algorithm generates minimal network with lesser computation time to achieve higher classification performance.

AB - This paper addresses sequential learning algorithm for self-adaptive resource allocation network classifier. Our approach makes use of self-adaptive error based control parameters to alter the training data sequence, evolve the network architecture, and learn the network parameters. In addition, the algorithm removes the training samples which are similar to the stored knowledge in the network. Thereby, it avoids the over-training problem and reduces the training time significantly. Use of misclassification information and hinge loss error in growing/learning criterion helps in approximating the decision function accurately. The performance evaluation using balanced and imbalanced data sets shows that the proposed algorithm generates minimal network with lesser computation time to achieve higher classification performance.

KW - Extended Kalman filter

KW - Multi-category classification

KW - Resource allocation network

KW - Self-adaptive control parameters

KW - Sequential learning

UR - http://www.scopus.com/inward/record.url?scp=78649975292&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78649975292&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2010.07.003

DO - 10.1016/j.neucom.2010.07.003

M3 - Article

VL - 73

SP - 3012

EP - 3019

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 16-18

ER -