National Changhua University of Education Institutional Repository : Item 987654321/15979

English | 正體中文 | 简体中文 | Items with full text/Total items : 6507/11669
Visitors : 30030608 Online Users : 530

RC Version 3.2 © Powered By DSPACE, MIT. Enhanced by NTU Library IR team.

Scope Adv. Search

Login ‧ Upload ‧ Help ‧ About ‧ Administer

NCUEIR > College of Management > Department of Information Management > Proceedings > Item 987654321/15979

Please use this identifier to cite or link to this item: http://ir.ncue.edu.tw/ir/handle/987654321/15979

Title:	On the Parallelization and Optimization of the Genetic-Based ANN Classifier for the Diagnosis of Students with Learning Disabilities
Authors:	Wu, Tung-Kuang;Huang, Shian-Chang;Meng, Ying-Ru;Lin, Ya-Li;Chang, Hsu
Contributors:	資訊管理學系
Keywords:	Learning Disabilities;Genetic Algorithm;Evolutionary Computation;Grid Computing;Neural Network
Date:	2010-10
Issue Date:	2013-04-22T07:37:30Z
Publisher:	IEEE
Abstract:	Diagnosis of students with learning disabilities has long been a difficult issue as it requires extensive man power and takes a long time. Through genetic algorithm based feature selection method and genetic based parameters optimization, artificial neural network (ANN) classifier has proven to be a good predictor to the diagnosis of students with learning disabilities. In this study, we keep focusing on the ANN model and compare three strategies of parallelizing the ANN parameter optimization procedure with OpenMP and MPI APIs. Not surprisingly, the outcomes show that all three parameter optimization procedures indeed converged or executed more quickly with the aid of parallel processing. In particular, the genetic-based method tends to derive the best accuracy and require less execution time. Most important of all, potentially due to a more diverse search space provided by the distributed parallel processing environment, the accuracy of the genetic-based ANN classifier may also be improved in general. In addition, with appropriate combinations of features and parameters setting, the accuracy in LD identification model has exceeded the 90% mark (using 5-fold cross validation), which is the best we have achieved so far. The result suggests that genetic-based (or perhaps similar) optimization methods may be benefited, both in reducing execution time and achieving better outcome, from current gridbased computing technologies.
Relation:	2010 IEEE Conference on Systems, Man and Cybernetics (SMC2010), : 4263-4269
Appears in Collections:	[Department of Information Management] Proceedings

Files in This Item:

File	Size	Format
index.html	0Kb	HTML	650	View/Open

Loading...