多子超球支持向量机算法研究  

Research on Multiple Sub-Hyper-Sphere Support Vector Machine

作　　者：陈鹏[1] 刘爽[2] 李灵华[2] 

机构地区：[1]大连东软信息学院计算机科学与技术系,辽宁大连116023 [2]大连民族学院计算机科学与工程学院,辽宁大连116600

出　　处：《微电子学与计算机》2014年第12期28-33,共6页Microelectronics & Computer

基　　金：国家自然科学基金资助项目(71303031)

摘　　要：提出了一种多子超球支持向量机算法.首先计算所有类别的分类超球,然后统计所有超球的位置关系,记入相应的数据集合.对相交集合中的超球,通过键值索引映射计算其交叉系数,根据该系数选择相交类别数据差异度最大的超球划分子超球,子超球个数由交叉系数确定.实验结果表明提出的算法在提高分类精度的同时,降低了决策规则求解的复杂度.This paper put forward a multiple sub-hyper-sphere support vector machine.New algorithm computed hyper-spheres for all classes first,and then obtained position relationships between hyper-spheres and saved them in the corresponding data set.For hyper-spheres in the intersection set,it computed overlap coefficient based on map of key value index,and partitioned the hyper-sphere with maximized difference degree into sub-hyper-spheres.Number of sub-hyper-spheres is based on the overlap coefficient.With experimental results compared to other hyper-sphere support vector machines,new proposed algorithm improves the performance of the resulting classifier and decreases computation complexity for decision.

关 键 词：超球支持向量机 多分类 相交 交叉系数 差异度 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]