Title :
An SMO Approach to Fast SVM for Classification of Large Scale Data
Author :
Juanxi Lin ; Mengnan Song ; Jinglu Hu
Author_Institution :
Grad. Sch. of Inf., Production & Syst., Waseda Univ., Kitakyushu, Japan
Abstract :
In this paper, a novel approach is proposed as a new fast Support Vector Machines (SVM) basing on sequential minimal optimization(SMO), minimum enclosing ball(MEB) approach and active set strategy. The combination with these 3 techniques largely accelerates the training process of SVM, attains fewer support vectors(SVs) as well as obtains a acceptable accuracy comparing to original SVM. From simulation results, it is stated that the proposed method will be a good alternative for classification of large scale data.
Keywords :
optimisation; pattern classification; support vector machines; MEB approach; SMO approach; active set strategy; fast SVM; large scale data classification; minimum enclosing ball approach; sequential minimal optimization; support vector machine; Accuracy; Educational institutions; Kernel; Production; Simulation; Support vector machines; Training;
Conference_Titel :
IT Convergence and Security (ICITCS), 2014 International Conference on
Conference_Location :
Beijing
DOI :
10.1109/ICITCS.2014.7021735
