首页 | 本学科首页   官方微博 | 高级检索  
     

基于壳向量的线性支持向量机快速增量学习算法
引用本文:李东晖,杜树新,吴铁军.基于壳向量的线性支持向量机快速增量学习算法[J].浙江大学学报(自然科学版 ),2006,40(2):202-206.
作者姓名:李东晖  杜树新  吴铁军
作者单位:浙江大学 工业控制技术国家重点实验室,浙江 杭州 310027
摘    要:提出了一种新的基于壳向量的增量式支持向量机快速学习算法.在增量学习的过程中,利用训练样本集中的几何信息,在样本中选取一部分最有可能成为支持向量的样本--壳向量,它是支持向量集的一个规模较小的扩展集,将其作为新的训练样本集,再进行支持向量训练.这在很大程度上减少了求取支持向量过程中的二次优化运算时间,使增量学习的训练速度大为提高.与单纯使用支持向量代表样本数据集合进行增量学习的传统算法相比,使用该算法使分类精度得到了提高.针对肝功能检测标准数据集(BUPA)的实验验证了该算法的有效性.

关 键 词:增量算法  支持向量机  壳向量
文章编号:1008-973X(2006)02-0202-05
收稿时间:2005-02-23
修稿时间:2005-02-23

Fast incremental learning algorithm of linear support vector machine based on hull vectors
LI Dong-hui,DU Shu-xin,WU Tie-jun.Fast incremental learning algorithm of linear support vector machine based on hull vectors[J].Journal of Zhejiang University(Engineering Science),2006,40(2):202-206.
Authors:LI Dong-hui  DU Shu-xin  WU Tie-jun
Affiliation:National Key Laboratory of Industrial Control Technology, Zhejiang University, Hangzhou 310027, China
Abstract:A new geometric fast incremental learning algorithm for support vector machines(SVM) was proposed.A set of hull vectors most likely to become the support vectors are extracted from the training samples by using the geometric information in these samples.In the incremental learning process,the obtained hull vector set and a new sample set are conjoined as the updated training sample set,which greatly reduces the time consumed in solving sequential quadratic optimization problems in incremental SVM training and speeds up the training process.Compared with the existing incremental SVM learning algorithms in which only support vectors are used to represent the original sample set,the proposed algorithm improves the classification precision.Experiments based on a standard BUPA dataset in liver function tests validated the effectiveness of the algorithm.
Keywords:incremental algorithm  support vector machine  hull vector
本文献已被 CNKI 维普 万方数据 等数据库收录!
点击此处可从《浙江大学学报(自然科学版 )》浏览原始摘要信息
点击此处可从《浙江大学学报(自然科学版 )》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号