Incremental support vector machine algorithm based on multi-kernel learning

Abstract

Abstract: A new incremental support vector machine (SVM) al-gorithm is proposed which is based on multiple kernel learning.Through introducing multiple kernel learning into the SVM incre-mental learning, large scale data set learning problem can besolved effectively. Furthermore, different punishments are adoptedin allusion to the training subset and the acquired support vectors,which may help to improve the performance of SVM. Simulationresults indicate that the proposed algorithm can not only solve themodel selection problem in SVM incremental learning, but alsoimprove the classification or prediction precision. Keywords: support vector machine (SVM), incremental learning,multiple kernel learning (MKL). DOI: 10.3969/j.issn.1004-4132.2011.04.021 1. Introduction Supportvectormachines(SVMs),proposedbyVapnik[1],areabletotranslatethenonlinearproblemintolinearprob-lemandobtaintheglobaloptimalsolutionbyfollowingthestructure risk minimization and integrating manifold tech-niquessuchasmaximalmarginhyper-plane,Mercerkerneland slack variables. In recent years, SVMs are extensivelyapplied to pattern recognition, system modeling, time se-ries predictionand otherfields [2,3].Unfortunately, SVMs do not scale well with respectto the size of training data. Given