Please use this identifier to cite or link to this item:
Full metadata record
DC FieldValueLanguage
dc.contributor.authorPapangkorn Inkeawen_US
dc.contributor.authorJakramate Bootkrajangen_US
dc.contributor.authorTeresa Gonçalvesen_US
dc.contributor.authorJeerayut Chaijaruwanichen_US
dc.description.abstract© 2018, Springer Nature Switzerland AG. Constructing a handwritten character recognition model is considered challenging partly due to the high variety of handwriting styles and the limited amount of training data. In practice, only a handful of labeled examples from limited number of writers are provided during the training of the model. Still, a large collection of already available unlabeled handwritten character data from several sources are often left unused. To alleviate the problem of small training sample size, we propose a graph-based active semi-supervised learning approach for handwritten character recognizer construction. The method iteratively builds a neighborhood graph of all examples including the unlabeled ones, assigns pseudo labels to the unlabeled data and retrains the model. Additionally, the label of the least confident pseudo label according to a newly proposed uncertainty measure is to be requested from the oracle. Experiments on NIST handwritten digits dataset demonstrated that the proposed learning method better utilizes the unlabeled data compared to existing approaches as measured by recognition accuracy. In addition, our active learning strategy is also more effective compared to baseline strategies.en_US
dc.subjectComputer Scienceen_US
dc.titleHandwritten Character Recognition Using Active Semi-supervised Learningen_US
dc.typeBook Seriesen_US
article.title.sourcetitleLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)en_US
article.volume11314 LNCSen_US Mai Universityen_US of Évoraen_US
Appears in Collections:CMUL: Journal Articles

Files in This Item:
There are no files associated with this item.

Items in CMUIR are protected by copyright, with all rights reserved, unless otherwise indicated.