计算机应用与软件2012，Vol.29Issue(4)：121-124,4.
核空间结合样本中心角度的支持向量机增量算法
INCREMENTAL SVM ALGORITHM BASED ON COMBINATION OF KERNEL SPACE AND SAMPLE CENTRE ANGLE
摘要
Abstract
To improve the training accuracy of incremental algorithm,in kernel-induced feature spaces,we firstly get two kind of centres in original training set and the ultra-normal plane of two centres,then obtain the ratios of the distances between original training set and the ultra-normal plane and between that and the midpoint of two centres, and finally generate the new training set by combining n sample points with smallest ratios with the support vectors in original training set and the samples in incremental set which contravenes KKT conditions. The presented mathematical model in end of the paper shows that the algorithm does not need to calculate the kernel-induced feature spaces, it retains more numbers of support vector than existing incremental support vector machine algorithms and ensures the training accuracy.关键词
支持向量机/KKT/增量算法/核空间/超平面/样本中心Key words
SVM/KKT/Incremental algorithm/Kernel space/Hyperplane/Sample centre分类
信息技术与安全科学引用本文复制引用
夏书银,王越,张权..核空间结合样本中心角度的支持向量机增量算法[J].计算机应用与软件,2012,29(4):121-124,4.基金项目
重庆市科委攻关项目(CSTC,2009AC208). （CSTC,2009AC208）
