LS-SVM applied for Photometric Classification of Quasars and Stars

Zhang, Yanxia

The major demerit of Support vector machines (SVM) is its higher computational cost for a quadratic programming (QP) problem. In order to overcome this problem, Least Squares Support Vector Machines (LS-SVM) is put forward. LS-SVM's solution is given by a linear system, which makes SVM method more generally simple and applicable. In this paper, LS-SVM is used for classification of quasars and stars from SDSS DR7 photometric database. The result shows that LS-SVM is highly efficient and powerful especially for large scale problem and has comparable performance with that of SVM.

Return to poster list