Study on parameters selection of LSSVR based on grid-diamond search method

Author

Hou, Li-kun ; Yang, Qing-Xin

Author_Institution

Province-Minist. Joint Key Lab. of Electromagn. Field & Electr. Apparatus Reliability, Hebei Univ. of Technol., Tianjin, China

Volume

2

fYear

2009

fDate

12-15 July 2009

Firstpage

1219

Lastpage

1224

Abstract

Determining the kernel function and regularization parameters for support vector machine (SVM) is very problem dependent in practice. A popular method to deciding the kernel parameters is cross validation method. But this makes the training process time consuming. In this paper we propose using grid diamond search method to choose the kernel parameters. Experiment results show that the grid-diamond search method can choose proper parameters of LSSVR and GDS is the fastest algorithm among the selecting parameter while providing better simulation result.

Keywords

least squares approximations; parameter estimation; regression analysis; search problems; support vector machines; cross validation method; grid diamond search; kernel parameter; least square version; support vector machine; Cybernetics; Electromagnetic fields; Equations; Kernel; Least squares methods; Machine learning; Quadratic programming; Search methods; Support vector machine classification; Support vector machines; Diamond searching (DS); Grid search; Kernel function; Least Square SVR; Parameters selection;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning and Cybernetics, 2009 International Conference on

Conference_Location

Baoding

Print_ISBN

978-1-4244-3702-3

Electronic_ISBN

978-1-4244-3703-0

Type

conf

DOI

10.1109/ICMLC.2009.5212375

Filename

5212375