支持向量机在模式识别中的核函数特性分析  

Support vector machine and kernel function characteristic analysis in pattern recognition

作　　者：李盼池[1] 许少华[1] 

机构地区：[1]大庆石油学院计算机科学与工程学院,黑龙江大庆163318

出　　处：《计算机工程与设计》2005年第2期302-304,共3页Computer Engineering and Design

基　　金：国家自然科学基金项目(60373102)。

摘　　要：支持向量机是20世纪90年代中期发展起来的一种机器学习技术,与传统人工神经网络不同之处在于前者基于结构风险最小化原理,后者基于经验风险最小化原理。支持向量机不仅结构简单,而且技术性能尤其是泛化能力与BP神经网络相比有明显提高。讨论了支持向量机的分类原理,并用多项式函数、径向基函数和感知机函数等3种核函数作为内积回旋,分别以平面点集分类、手写体汉字识别及双螺旋线识别为例,在不同的结构参数下进行了仿真实验,并对3种核函数的分类特性进行了对比分析,给出了在不同模式识别问题中3种核函数的选择条件。Support vector machine (SVM) is a method of machine learning developed in the middle period in 1990s. The difference between SVM and neural network is that the former is based on structure risk minimization principle in pattern recognition, and the latter is based on experience risk minimization principle in pattern recognition. SVM has not only simpler structure, but also better performance, especially better generalization ability. The classification principle of SVM is discussed, and three kinds of kernel function are applied to SVM, which are multinomial function, radial basis function, perceptron function. Finally comparison of three kernel function characteristic is presented by three recognition examples: classification of point set in plane, recognition of similar script Chinese characters, and recognition of double screw curve. The rules of kernel function selection in different recognition matter are presented.

关 键 词：支持向量机 核函数 模式识别 感知机 手写体汉字识别 机器学习 结构风险最小化 内积 平面点集 多项式函数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]