摘要
阐述矩阵能极分解生成一个对称半正定核矩阵的特性,对RBF核进行极分解,并结合多项式核,构造了一个性能很好的混合核函数。在CEMENT数据集和IRIS数据集上进行实验,采用基于极分解下的混合核,来与RBF核进行了比较。结果表明,使用该核的SVM,其支持向量的个数少、分类错误低、并有较高的训练速度。还进一步发现,该混合核有效地抑制了局部核函数RBF所引起的预测输出波动。
This paper describes a method of mixtures of kernels based on extremum disassociation and polynomial kenels.The experiments on the CEMENT data and the IRIS data is used to enable the mixtures of Kernels to compare with RBF kenels,which shows SVM using the mixtures of Kernels reduces the number of the support vector and bring fine classification result,and whatsmore,improves the trainning speed.It is further found,the mixtures of kenels availablyrestrain the RBF kenel fluctuation based on loacal kenels of RBF kenerls.
作者
许鑫
XU Xin(Taizhou City Urban Social Governance Modernization Command Center,Jiangsu 225300,China)
出处
《电子技术(上海)》
2023年第12期22-24,共3页
Electronic Technology
关键词
支持向量机
混合核函数
极分解
局部核
support vector machine(SVM)
mixtures of kernels
extremum disassociation
local kenels