支持向量机核函数的构造方法研究与分析  

Research on structure of support vector machine kernel function

作　　者：杨冬云[1] 李数函[2] 

机构地区：[1]黑龙江工程学院电子工程系,黑龙江哈尔滨150050 [2]成都市技师学院电子信息工程系,四川成都610031

出　　处：《高师理科学刊》2010年第2期23-26,共4页Journal of Science of Teachers＇College and University

基　　金：黑龙江省教育厅2007年度科学技术研究项目(11521227)

摘　　要：支持向量机是近年来发展的以统计学习理论为基础的新型学习机,核函数的选择无疑是其重要的核心问题.先对核函数进行定义,然后通过对核理论进行性质分析,最后明确了对最优核函数的选择与构造,从理论上为核函数的选择提供了参考.该学习机用结构风险代替经验风险,因而具有传统的神经网络无法相比的优势.Support vector machine is a newly developed learning machine based on statistical learning theory in recent years. The choice of kernel function is undoubtedly the core issue of utmost importance. At first the definition of kernel function was given, and then the nature of kernel theory was analyzed, finally the optimal choice of kernel function and its structure was clarified, which provides a theoretical reference on the choice of kernel function. This learning machine uses structural risk instead of experience risk thus owns an uncomparable advantage over the traditional neural network.

关 键 词：支持向量机 核函数 核理论 最优核函数 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]