基于连续过松弛方法的支持向量回归算法.pdfVIP

1000-9825/2004/15(02)0200 ©2004 Journal of Software 软 件 学 报 Vol.15, No.2 ∗ 基于连续过松弛方法的支持向量回归算法 + 全 勇 , 杨 杰, 姚莉秀, 叶晨洲 (上海交通大学 图像处理及模式识别研究所,上海 200030) Successive Overrelaxation for Support Vector Regression + QUAN Yong , YANG Jie, YAO Li-Xiu, YE Chen-Zhou (Institute of Image Processing and Pattern Recognition, Shanghai Jiaotong University, Shanghai 200030, China) + Corresponding author: Phn: +86-10 E-mail: quanysjtu@ Received 2002-12-08; Accepted 2003-04-15 Quan Y, Yang J, Yao LX, Ye CZ. Successive overrelaxation for support vector regression. Journal of Software, 2004,15(2):200~206. /1000-9825/15/200.htm Abstract: Training a SVR (support vector regression) requires the solution of a very large QP (quadratic programming) optimization problem. Despite the fact that this type of problem is well understood, the existing training algorithms are very complex and slow. In order to solve these problems, this paper firstly introduces a new way to make SVR have the similar mathematic form as that of a support vector machine. Then a versatile iterative method, successive overrelaxation, is proposed. Experimental results show that this new method converges considerably faster than other methods that require the presence of a substantial amount of data in memory. The results give guidelines for the application of this method to large domains. Key words: support vector regression; support vector machine; successive overrelaxation; quadratic programming; chunking algorithm 摘 要: 支持向量回归(support vector regression,简称SVR)训练算法需要解决在大规模样本条件下的凸二次规划 (quadratic programming