Iteratively reweighted least squares classifier and its l2- and l1-regularized Kernel versions

Bulletin of the Polish Academy of Sciences. Technical Sciences

J. ??skiInstitute of Electronics, Silesian University of Technology, 16 Akademicka St., 44-100 Gliwice, PolandOther articles by this author:De Gruyter OnlineGoogle Scholar¹ 

[1] Institute of Electronics, Silesian University of Technology, 16 Akademicka St., 44-100 Gliwice, Poland

关键词: Keywords: classifier design;    IRLS;    conjugate gradient optimization;    gradient projection;    Kernel matrix;   

DOI  :  10.2478/v10175-010-0018-2

学科分类：工程和技术（综合）

来源: Polska Akademia Nauk * Centrum Upowszechniania Nauki / Polish Academy of Sciences, Center for the Advancement of Science

【 摘 要 】

This paper introduces a new classifier design method based on regularized iteratively reweighted least squares criterion function. The proposed method uses various approximations of misclassification error, including: linear, sigmoidal, Huber and logarithmic. Using the represented theorem a kernel version of classifier design method is introduced. The conjugate gradient algorithm is used to minimize the proposed criterion function. Furthermore, l1-regularized Kernel versions of the classifier is introduced. In this case, the gradient projection is used to optimize the criterion function. Finally, an extensive experimental analysis on 14 benchmark datasets is given to demonstrate the validity of the introduced methods.

【 授权许可】

Unknown