Tuning Recurrent Neural Networks for Recognizing Handwritten Arabic Words

Handle URI:
http://hdl.handle.net/10754/550837
Title:
Tuning Recurrent Neural Networks for Recognizing Handwritten Arabic Words
Authors:
Qaralleh, Esam; Abandah, Gheith; Jamour, Fuad Tarek
Abstract:
Artificial neural networks have the abilities to learn by example and are capable of solving problems that are hard to solve using ordinary rule-based programming. They have many design parameters that affect their performance such as the number and sizes of the hidden layers. Large sizes are slow and small sizes are generally not accurate. Tuning the neural network size is a hard task because the design space is often large and training is often a long process. We use design of experiments techniques to tune the recurrent neural network used in an Arabic handwriting recognition system. We show that best results are achieved with three hidden layers and two subsampling layers. To tune the sizes of these five layers, we use fractional factorial experiment design to limit the number of experiments to a feasible number. Moreover, we replicate the experiment configuration multiple times to overcome the randomness in the training process. The accuracy and time measurements are analyzed and modeled. The two models are then used to locate network sizes that are on the Pareto optimal frontier. The approach described in this paper reduces the label error from 26.2% to 19.8%.
KAUST Department:
Computer Science Program
Citation:
Tuning Recurrent Neural Networks for Recognizing Handwritten Arabic Words 2013, 06 (10):533 Journal of Software Engineering and Applications
Publisher:
Scientific Research Publishing, Inc,
Journal:
Journal of Software Engineering and Applications
Issue Date:
Oct-2013
DOI:
10.4236/jsea.2013.610064
Type:
Article
ISSN:
1945-3116; 1945-3124
Additional Links:
http://www.scirp.org/journal/PaperDownload.aspx?DOI=10.4236/jsea.2013.610064
Appears in Collections:
Articles; Computer Science Program

Full metadata record

DC FieldValue Language
dc.contributor.authorQaralleh, Esamen
dc.contributor.authorAbandah, Gheithen
dc.contributor.authorJamour, Fuad Tareken
dc.date.accessioned2015-04-28T11:58:42Zen
dc.date.available2015-04-28T11:58:42Zen
dc.date.issued2013-10en
dc.identifier.citationTuning Recurrent Neural Networks for Recognizing Handwritten Arabic Words 2013, 06 (10):533 Journal of Software Engineering and Applicationsen
dc.identifier.issn1945-3116en
dc.identifier.issn1945-3124en
dc.identifier.doi10.4236/jsea.2013.610064en
dc.identifier.urihttp://hdl.handle.net/10754/550837en
dc.description.abstractArtificial neural networks have the abilities to learn by example and are capable of solving problems that are hard to solve using ordinary rule-based programming. They have many design parameters that affect their performance such as the number and sizes of the hidden layers. Large sizes are slow and small sizes are generally not accurate. Tuning the neural network size is a hard task because the design space is often large and training is often a long process. We use design of experiments techniques to tune the recurrent neural network used in an Arabic handwriting recognition system. We show that best results are achieved with three hidden layers and two subsampling layers. To tune the sizes of these five layers, we use fractional factorial experiment design to limit the number of experiments to a feasible number. Moreover, we replicate the experiment configuration multiple times to overcome the randomness in the training process. The accuracy and time measurements are analyzed and modeled. The two models are then used to locate network sizes that are on the Pareto optimal frontier. The approach described in this paper reduces the label error from 26.2% to 19.8%.en
dc.publisherScientific Research Publishing, Inc,en
dc.relation.urlhttp://www.scirp.org/journal/PaperDownload.aspx?DOI=10.4236/jsea.2013.610064en
dc.rightsThis is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.en
dc.subjectOptical Character Recognitionen
dc.subjectHandwritten Arabic Wordsen
dc.subjectRecurrent Neural Networksen
dc.subjectDesign of Experimentsen
dc.titleTuning Recurrent Neural Networks for Recognizing Handwritten Arabic Wordsen
dc.typeArticleen
dc.contributor.departmentComputer Science Programen
dc.identifier.journalJournal of Software Engineering and Applicationsen
dc.eprint.versionPublisher's Version/PDFen
dc.contributor.institutionComputer Engineering Department, Princess Sumaya University for Technology, Amman, Jordanen
dc.contributor.institutionComputer Engineering Department, The University of Jordan, Amman, Jordanen
kaust.authorJamour, Fuad Tareken
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.