A condensed polynomial neural network for classification using swarm intelligence

S. Dehuri, B. B. Misra, A. Ghosh, Sung-Bae Cho

Research output: Contribution to journalArticle

11 Citations (Scopus)

Abstract

A novel condensed polynomial neural network using particle swarm optimization (PSO) technique is proposed for the task of classification in this paper. In solving classification task classical algorithms such as polynomial neural network (PNN) and its variants need more computational time as the partial descriptions (PDs) grow over the training period layer-by-layer and make the network very complex. Unlike PNN the proposed network needs to generate the partial description for a single layer. The discrete PSO (DPSO) is used to select a relevant set of PDs as well as features with a hope to get better accuracy, which are in turn fed to the output neuron. The weights associated with the links from hidden to output neuron is optimized by PSO for continuous domain (CPSO). Performance of this model is compared with the results obtained from PNN. Simulation result shows that the performance of this model both in processing time and accuracy, is encouraging for harnessing its power in domain with large and complex data particularly in data mining area.

Original languageEnglish
Pages (from-to)3106-3113
Number of pages8
JournalApplied Soft Computing Journal
Volume11
Issue number3
DOIs
Publication statusPublished - 2011 Apr 1

Fingerprint

Polynomials
Particle swarm optimization (PSO)
Neural networks
Neurons
Data mining
Swarm intelligence
Processing

All Science Journal Classification (ASJC) codes

  • Software

Cite this

Dehuri, S. ; Misra, B. B. ; Ghosh, A. ; Cho, Sung-Bae. / A condensed polynomial neural network for classification using swarm intelligence. In: Applied Soft Computing Journal. 2011 ; Vol. 11, No. 3. pp. 3106-3113.
@article{7b4a884f0cb8484a9ffbb2d1bb5b8bd5,
title = "A condensed polynomial neural network for classification using swarm intelligence",
abstract = "A novel condensed polynomial neural network using particle swarm optimization (PSO) technique is proposed for the task of classification in this paper. In solving classification task classical algorithms such as polynomial neural network (PNN) and its variants need more computational time as the partial descriptions (PDs) grow over the training period layer-by-layer and make the network very complex. Unlike PNN the proposed network needs to generate the partial description for a single layer. The discrete PSO (DPSO) is used to select a relevant set of PDs as well as features with a hope to get better accuracy, which are in turn fed to the output neuron. The weights associated with the links from hidden to output neuron is optimized by PSO for continuous domain (CPSO). Performance of this model is compared with the results obtained from PNN. Simulation result shows that the performance of this model both in processing time and accuracy, is encouraging for harnessing its power in domain with large and complex data particularly in data mining area.",
author = "S. Dehuri and Misra, {B. B.} and A. Ghosh and Sung-Bae Cho",
year = "2011",
month = "4",
day = "1",
doi = "10.1016/j.asoc.2010.12.013",
language = "English",
volume = "11",
pages = "3106--3113",
journal = "Applied Soft Computing",
issn = "1568-4946",
publisher = "Elsevier BV",
number = "3",

}

A condensed polynomial neural network for classification using swarm intelligence. / Dehuri, S.; Misra, B. B.; Ghosh, A.; Cho, Sung-Bae.

In: Applied Soft Computing Journal, Vol. 11, No. 3, 01.04.2011, p. 3106-3113.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A condensed polynomial neural network for classification using swarm intelligence

AU - Dehuri, S.

AU - Misra, B. B.

AU - Ghosh, A.

AU - Cho, Sung-Bae

PY - 2011/4/1

Y1 - 2011/4/1

N2 - A novel condensed polynomial neural network using particle swarm optimization (PSO) technique is proposed for the task of classification in this paper. In solving classification task classical algorithms such as polynomial neural network (PNN) and its variants need more computational time as the partial descriptions (PDs) grow over the training period layer-by-layer and make the network very complex. Unlike PNN the proposed network needs to generate the partial description for a single layer. The discrete PSO (DPSO) is used to select a relevant set of PDs as well as features with a hope to get better accuracy, which are in turn fed to the output neuron. The weights associated with the links from hidden to output neuron is optimized by PSO for continuous domain (CPSO). Performance of this model is compared with the results obtained from PNN. Simulation result shows that the performance of this model both in processing time and accuracy, is encouraging for harnessing its power in domain with large and complex data particularly in data mining area.

AB - A novel condensed polynomial neural network using particle swarm optimization (PSO) technique is proposed for the task of classification in this paper. In solving classification task classical algorithms such as polynomial neural network (PNN) and its variants need more computational time as the partial descriptions (PDs) grow over the training period layer-by-layer and make the network very complex. Unlike PNN the proposed network needs to generate the partial description for a single layer. The discrete PSO (DPSO) is used to select a relevant set of PDs as well as features with a hope to get better accuracy, which are in turn fed to the output neuron. The weights associated with the links from hidden to output neuron is optimized by PSO for continuous domain (CPSO). Performance of this model is compared with the results obtained from PNN. Simulation result shows that the performance of this model both in processing time and accuracy, is encouraging for harnessing its power in domain with large and complex data particularly in data mining area.

UR - http://www.scopus.com/inward/record.url?scp=79951854536&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79951854536&partnerID=8YFLogxK

U2 - 10.1016/j.asoc.2010.12.013

DO - 10.1016/j.asoc.2010.12.013

M3 - Article

AN - SCOPUS:79951854536

VL - 11

SP - 3106

EP - 3113

JO - Applied Soft Computing

JF - Applied Soft Computing

SN - 1568-4946

IS - 3

ER -