计算机工程 ›› 2012, Vol. 38 ›› Issue (12): 185-187.doi: 10.3969/j.issn.1000-3428.2012.12.055

• 人工智能及识别技术 • 上一篇    下一篇

L1范数正则化SVM聚类算法

刘建伟,李双成,付 捷,罗雄麟   

  1. (中国石油大学自动化研究所,北京 102249)
  • 收稿日期:2011-08-01 出版日期:2012-06-20 发布日期:2012-06-20
  • 作者简介:刘建伟(1966-),男,副研究员、博士,主研方向:机器学习;李双成、付 捷,硕士研究生;罗雄麟,教授、博士
  • 基金项目:

    中国石油大学(北京)基础学科研究基金资助项目

L1-norm Regularized SVM Clustering Algorithm

LIU Jian-wei, LI Shuang-cheng, FU Jie, LUO Xiong-lin   

  1. (Research Institute of Automation, China University of Petroleum, Beijing 102249, China)
  • Received:2011-08-01 Online:2012-06-20 Published:2012-06-20

摘要: 提出L1范数正则化支持向量机(SVM)聚类算法。该算法能够同时实现聚类和特征选择功能。给出L1范数正则化SVM聚类原问题和对偶问题形式,采用类似迭代坐标下降的方法求解困难的混合整数规划问题。在多组数据集上的实验结果表明,L1范数正则化SVM聚类算法聚类准确率与L2范数正则化SVM聚类算法相近,而且能够实现特征选择。

关键词: 支持向量机, L1范数, 正则化, 特征选择, 聚类, 对偶问题

Abstract: This paper proposes a L1-norm regularized Support Vector Machine(SVM) clustering algorithm. The proposed clustering algorithm can fulfill simultaneous clustering construction and feature selection. Primal and dual form of L1-norm regularized SVM clustering problem is introduced, and an algorithm that iterates coodinate-wise desent approach is adopted to solve difficult mixed integer programming. Experimental results on real datasets show that the predict accuracy of the presented algorithm is compared with L2-norm regularized SVM clustering algorithm, and can take feature selection.

Key words: Support Vector Machine(SVM), L1-norm, regularization, feature selection, clustering, dual problem

中图分类号: