DocumentCode
1694699
Title
Research on ECOC SVMs
Author
Yan, Zhigang
Author_Institution
Key Lab. for Land Environ. & Disaster Monitoring of SBSM, China Univ. of Min. & Technol., Xuzhou, China
fYear
2010
Firstpage
2838
Lastpage
2842
Abstract
The generalization performance of ECOC SVMs was analyzed, as a result, the performance of ECOC SVMs is mainly determined by the SVM classifiers corresponding to its codewords, litttlely by its mathematical characteristics. The performance of each SVM is ordered by their separating margins and cross-validation error ratios. Three types of ECOC SVMs, whose performance are better, worse and common, are constructed by selecting different set of SVMs, the ECOC SVMs with a better performance might be constructed by the set of SVMs whose performances are better too, otherwise, its performance might be worse, which supports our viewpoint effectively and points out the direction for improving ECOC SVMs.
Keywords
error correction codes; pattern classification; support vector machines; ECOC SVM; SVM classifiers; error correcting output codes; support vector machine; Informatics; Learning; Machine learning; Monitoring; Pattern recognition; Support vector machines; ECOC SVMs; Generalization Performance; SVM;
fLanguage
English
Publisher
ieee
Conference_Titel
Intelligent Control and Automation (WCICA), 2010 8th World Congress on
Conference_Location
Jinan
Print_ISBN
978-1-4244-6712-9
Type
conf
DOI
10.1109/WCICA.2010.5554718
Filename
5554718
Link To Document