回归支持向量机的改进序列最小优化学习算法-软件学报.PDF

1000-9825/2003/14(12)2006 ©2003 Journal of Software 软 件 学 报 Vol.14, No.12 ∗ 回归支持向量机的改进序列最小优化学习算法 + 张浩然 , 韩正之 (上海交通大学 自动化系,上海 200030) An Improved Sequential Minimal Optimization Learning Algorithm for Regression Support Vector Machine + ZHANG Hao-Ran , HAN Zheng-Zhi (Department of Automation, Shanghai Jiaotong University, Shanghai 200030, China) + Corresponding author: Phn: 86-21ext 810, E-mail: gczhr@ Received 2002-11-04; Accepted 2003-03-04 Zhang HR, Han ZZ. An improved sequential minimal optimization learning algorithm for regression support vector machine. Journal of Software, 2003,14(12): 2006~2013. /1000-9825/14/2006.htm Abstract: Support vector machine (SVM) is a learning technique based on the structural risk minimization principle, and it is also a class of regression method with a good generalization ability. This paper presents an improved SMO (sequential minimal optimization) algorithm to train the regression SVM, which gives an the analytical solution to the QP problem of size two. A new working set selection method and a stopping condition are developed. The simulation results show that the improved SMO algorithm is significantly faster and more precise than the original SMO one. Key words: support vector machine; kernel method; regression; SMO (sequential minimal optimization) 摘 要: 支持向量机(support vector machine, 简称 SVM)是一种基于结构风险最小化原理的学习技术,也是一种新 的具有很好泛化性能的回归方法,提出了实现回归支持向量机的一种改进的 SMO(sequential minimal optimization) 算法,给出了两变量子优化问题的解析解,设计了新的工作集选择方法和停止条件,仿真实例说明,所提出的 SMO 算 法比原始 SMO 算法具有更快的运算速度. 关键词: 支持向量机;核方法;回归;序列最小优化 中图法分类号: TP18 文献标识码: A 自从 Vapnik[1]等人提出支持向量机(sup