Non-parametric Least Square Support Vector Machine

Author

Guo, Ning ; Chen, Xiankai ; Ma, Yingdong ; Chen, Gorge

Author_Institution

Center of Digital Media Comput., Shenzhen Inst. of Adv. Technol., Shenzhen, China

fYear

2012

fDate

29-31 May 2012

Firstpage

93

Lastpage

96

Abstract

Distinguished from Semi-Definite Programming (SDP) and Quadratically Constrained Quadratic Program (QCQP) which are classical solutions of multiple kernel learning, we apply Semi-Infinite Linear Program (SILP) to deal with multiple kernels learning in Least Square Support Vector Machine (LSSVM). Furthermore the regularization parameter is added as an extra variable to learn. This algorithm avoids the computational cost consuming by cross validation and make algorithm more convenient and practical.

Keywords

learning (artificial intelligence); least squares approximations; linear programming; quadratic programming; support vector machines; LSSVM; QCQP; SDP; SILP; cross validation; multiple kernel learning; nonparametric least square support vector machine; quadratically constrained quadratic program; regularization parameter; semidefinite programming; semiinfinite linear program; Accuracy; Classification algorithms; Kernel; Machine learning; Scalability; Support vector machines; Training; LSSVM; MKL; QCQP; SDP; SILP optimal solution;

fLanguage

English

Publisher

ieee

Conference_Titel

Natural Computation (ICNC), 2012 Eighth International Conference on

Conference_Location

Chongqing

ISSN

2157-9555

Print_ISBN

978-1-4577-2130-4

Type

conf

DOI

10.1109/ICNC.2012.6234737

Filename

6234737