A Linear One-Class Classifier based on 1-SVM
The One-Class Support Vector Machine (1-SVM) algorithm requires the solution of a quadratic programming problem regarding the dimension of the sample number. The complexity becomes highly intensive as the number of the data points increases. This paper proposes a novel algorithm named Linear OneClass Classifier (LOCC) based on 1-SVM. By enhancing the constraint inequations to equations and squaring the penalty term in Lagrangian, LOCC can transform the quadratic programming problem to a set of linear equations, which can be figured out analytically. The proposed algorithm is well applied to the hierarchical clustering and the kernel clustering method. Simulation results show that, compared with 1-SVM, the operation rate of LOCC improves greatly while its accuracy remains almost the same.
Lei Xu Guangzhou Zhao Liaoying Zhao
College of Electrical Engineering Zhejiang University Hangzhou, PR China, 310027 Institute of Computer Application Technology Hangzhou Dianzi University Hangzhou, PR China, 310018
国际会议
南宁
英文
2007-07-20（万方平台首次上网日期，不代表论文的发表时间）