摘要
多类分问题通常采用多个标准的二分类支持向量机来求解,在这种情况下,需要解多个二次规划问题.为了简化多类分类问题带来的计算复杂性,本文根据一类分类思想提出一种多类分类算法,所给算法通过引入核函数能够独立地对每一类样本形成一个紧致的优化区域,从而达到分类的目的.人工及实际数据库的仿真实验表明所给算法在保持良好的分类精度条件下,能有效降低程序的运行时间.
The multi-class classification problem is commonly solved by decomposition to several binary problems for which the standard Support Vector Machine(SVM)can be used. In this case, many quadratic programming problems need to be solved. In order to reduce computation complexity about multi-class classification problems, a new multi-class classification algorithm based on one-class classification idea is proposed. It can form a compact boundary about every single class sample by using kernel functions and accordingly obtain the aim of classification. Simulation is conducted on artificial and real database, which shows that the proposed method can guarantee the classification precision and reduce the running time of program.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2004年第3期357-361,共5页
Pattern Recognition and Artificial Intelligence
基金
广东省自然科学基金(No.032353)
关键词
支持向量机
多类分类
一类分类
核函数
Support Vector Machine
Multi-Class Classification
One-Class Classification
Kernel Function
