Online LS-SVM Learning for Classification Problems Based on Incremental Chunk

Zhifeng Hao, Shu Yu, Xiaowei Yang, Feng Zhao, Rong Hu, Yanchun Liang

Abstract: In this paper an online learning algorithm based on incremental chunk for LS-SVM (Least Square Support Vector Machines) classifiers is proposed. The training of the LS-SVM can be placed in a way of incremental chunk, which avoids computing large-scale matrix inverse but maintaining the precision when training and testing data. This online algorithm is especially useful for the large data set and practical applications where the data come in sequentially. Our experiments with four classification problems in UCI show that compared with LS-SVM, the computational cost of our algorithm is reduced obviously and the accuracy is retained.