一种新的支持向量机增量学习算法  

A New Incremental Learning for Support Vector Machine

作　　者：任广永[1] 

机构地区：[1]安庆师范学院计算机与信息学院,安徽安庆246011

出　　处：《河南科技大学学报（自然科学版）》2008年第4期54-57,共4页Journal of Henan University of Science And Technology:Natural Science

基　　金：国家自然科学基金项目(60773128)

摘　　要：支持向量机学习算法的本质是从训练集中寻找支持向量,因此能否通过训练算法能快速找出支持向量是衡量支持向量机算法优劣的重要标准。本文提出了一种新的快速训练支持向量机的增量学习算法,首先,给出边界向量的定义,然后,对一个给定的新加入的样本,新的学习方法验证其是否为边界向量,如果是,将其加入到训练集中重新训练支持向量机,如果不是,就舍弃,这样能达到减少训练样本、降低训练复杂性目的,最后,给出了一个增量学习算法。实验表明测试误差和支持向量数量与SMO算法大致相当,而训练速度明显加快。A new incremental learning method for support vector machine(SVM)is proposed,which trains SVM quickly and incrementally.In this paper,a definition of border vectors is given first.Border vectors are supported by vector candidates.Then for a given new-added sample,the proposed training method validates whether they are border vectors or not.If true,they are added to training sample set to retrain support vector machine,otherwise omitted.Hence,the training samples can be reduced and training complexity can be lessened.Finally,an incremental algorithm is presented to train SVM by using the selected border vectors.Experiment results show that the test error and support vector number of the proposed algorithm are almost the same as those of SMO algorithm.However,the training speed of the new incremental algorithm is more quickly than that of SMO method.

关 键 词：支持向量机 边界向量 增量学习 核函数 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]