基于向量集约简的精简支持向量机-软件学报.PDF

ISSN 1000-9825, CODEN RUXUEW E-mail: jos@ Journal of Software, Vol.18, No.11, November 2007, pp .2719−2727 DOI: 10. 1360/jos1 82719 Tel/Fax : +86-10 © 2007 by Journal of Software. All rights reserved. ∗ 基于向量集约简的精简支持向量机 + 曾志强, 高 济 (浙江大学 计算机科学与技术学院,浙江 杭州 310027) Simplified Support Vector Machine Based on Reduced Vector Set Method + ZENG Zhi-Qiang , GAO Ji (College of Computer Science and Technology, Zhejia ng University, Hangzhou 310027, China) + Corresponding author: E-mail: lbxzzq@163.com, Zeng ZQ, Gao J. Simplified support vector machine based on reduced vector set method. Journal of Software, 2007,18(11):2719−2727. /1000-9825/18/2719.htm Abstract: Existing works of reduced support vector set method find the reduced set vectors based on solving an unconstrained optimization problem with multivariables, which may suffer from numerical instability or get trapped in a local minimum. In this paper, a reduced set method relying on kernel-based clustering is presented to simplify SVM (support vector machine) solution. The method firstly organizes support vectors in clusters in feature space, and then, it finds the pre-images of the cluster centroids in feature space to construct a reduced vector set. This approach is conceptually simpler, involves only linear algebra and overcomes the difficulties existing in the former reduced set methods. Experimental results on real data sets indicate that the proposed method is effective in simplifying SVM solution while preserving machine’s generalization performance. Key words: support vector machine; reduced vector set; kernel-based clustering; pre-