计算机工程 ›› 2012, Vol. 38 ›› Issue (2): 175-177.doi: 10.3969/j.issn.1000-3428.2012.02.057

• 人工智能及识别技术 • 上一篇    下一篇

基于核熵成分分析的数据降维?

黄丽瑾,施 俊,钟 瑾   

  1. (上海大学通信与信息工程学院,上海 200072)
  • 收稿日期:2011-07-18 出版日期:2012-01-20 发布日期:2012-01-20
  • 作者简介:黄丽瑾(1986-),女,硕士研究生,主研方向:信号处理;施 俊,副教授;钟 瑾,硕士研究生
  • 基金项目:
    国家自然科学基金资助项目(60701021);上海市教育委员会科研创新基金资助项目(09YZ15);上海市教委重点学科建设基金资助项目(J50104);上海大学研究生创新基金资助项目(SHUCX112 137)

Data Dimension Reduction Based on Kernel Entropy Component Analysis

HUANG Li-jin, SHI Jun, ZHONG Jin   

  1. (School of Communication and Information Engineering, Shanghai University, Shanghai 200072, China)
  • Received:2011-07-18 Online:2012-01-20 Published:2012-01-20

摘要: 针对高维数据的维灾问题,采用核熵成分分析方法降维数据,并与主成分分析及核主成分分析方法进行对比。降维后的数据利用支持向量机算法进行分类,以验证算法有效性。实验结果表明,KECA在较低的维数时仍然能获得较好的分类精度,可以减少后续的处理复杂度和运行时间,适用于机器学习、模式识别等领域。

关键词: 降维, 核熵成分分析, 核主成分分析, 支持向量机

Abstract: Aiming at the curse of dimensionality, the kernel entropy component analysis(KECA) is used to reduce the dimension of data, which is compared with Principal Component Analysis(PCA) and Kernel PCA(KPCA). The low dimensional data after dimension reduction are classified by Support Vector Machine(SVM) algorithm to compare the accuracy. Experimental results indicate that high classification accuracy can be obtained at low dimension number with KECA, which reduces the processing complexity and running time. It suggests that KECA-based dimension reduction algorithm has the feasibility to be applied in the fields of machine learning, pattern recognition, etc.

Key words: dimension reduction, Kernel Entropy Component Analysis(KECA), Kernel Principal Component Analysis(KPCA), Support Vector Machine(SVM)

中图分类号: