一种改进的有向无环图支持向量机  

A Improved Directed Acyclic Graphs Support Vector Machine

作　　者：艾青[1] 秦玉平[2] 赵骥[1] 

机构地区：[1]辽宁科技大学软件学院 [2]渤海大学信息科学与工程学院

出　　处：《计算机工程与科学》2011年第10期145-148,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60603023);辽宁省教育厅资助科研课题(2010076)

摘　　要：构造合理的有向无环图是有向无环图支持向量机亟需解决的一个关键问题。本文提出一种改进的有向无环图支持向量机,根据超球支持向量机获得类的最小包围球,根据该最小包围球计算类与类之间的最短距离,根据该最短距离形成最短距离矩阵,根据该最短距离矩阵来构造有向无环图。实验结果表明,该改进算法较传统有向无环图支持向量机分类精度有明显提高。Constructing reasonable directed acyclic graphs is a key problem that is to be solved urgently. An improved directed acyclic graphs support vector machine is presented, which obtains the minimal hyper-spheres according to the hyper-sphere support vector machine, computes the minimal distances among classes according to the minimal hyper-spheres, gets the minimal distance matrix according to the minimal distances, and constructs the directed acyclic graphs according to the minimal distance matrix. The experimental results show that the algorithm has a higher classification precision, compared with the old directed acyclic graphs support vector machines.

关 键 词：有向无环图支持向量机(DAGSVM) 超球支持向量机 最短距离矩阵 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]