光滑支持向量机两种求解算法的比较  

Comparison between two solving methods for smooth support vector machine

作　　者：李广明[1] 刘群锋[1] 

机构地区：[1]东莞理工学院计算机学院,广东东莞523808

出　　处：《计算机应用》2009年第6期1612-1614,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60773050);广东省科技计划项目(2008B060600076)

摘　　要：BFGS-Armijo法和Newton-Armijo法是求解光滑模型的常用算法。分别列出用此两种算法求解光滑支持向量机模型(SSVM)的具体步骤,并用这两种算法对求解SSVM模型进行比较研究。数值实验结果表明:Newton-Armijo法的分类性能和BFGS-Armijo法基本相同,而分类效率比BFGS-Armijo法高出约26.2%。The BFGS-Armijo method and NewtonzArmijo method are two popular methods to solve the smooth models. The authors respectively listed the procedures of these two methods to solve Smooth Support Vector Machine （ SSVM）, and made a comparative study between the two methods to solve SSVM. The numerical results show that the classification performance of BFGS-Armijo method is almost the same as the Newton-Armijo method, but the classification efficiency of Newton-Armijo method is 26.2% higher than that of BFGS-Armijo method.

关 键 词：分类 支持向量机 Newton—Armijo法 BFGS—Armijo法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]