计算机科学 ›› 2015, Vol. 42 ›› Issue (4): 199-205.doi: 10.11896/j.issn.1002-137X.2015.04.040

• 人工智能 • 上一篇    下一篇

基于统计检验的核函数度量方法研究

王裴岩,蔡东风   

  1. 南京航空航天大学计算机科学与技术学院 南京210016;沈阳航空航天大学知识工程研究中心 沈阳110136,沈阳航空航天大学知识工程研究中心 沈阳110136
  • 出版日期:2018-11-14 发布日期:2018-11-14
  • 基金资助:
    本文受国家“十二五”科技支撑计划项目(2012BAH14F00)资助

Statistical Testing Based Research on Kernel Evaluation Measures

WANG Pei-yan and CAI Dong-feng   

  • Online:2018-11-14 Published:2018-11-14

摘要: 将统计检验方法应用于核函数度量。以核函数、规范化核函数、中心化核函数和核距离作为样本在特征空间中的几何关系度量,使用t检验和F检验等7种统计检验方法检验特征空间中同类样本间几何关系度量值与异类样本间几何关系度量值的分布差异,以此反映特征空间中同类样本间内聚性与异类样本间分离性间的差异。在11个UCI数据集上进行的核函数选择实验表明,基于统计检验的核度量方法达到或超过了核校准与特征空间核度量标准等方法的效果,适用于核函数度量;并且发现两类数据分布差异主要体现在了方差差异上。此外,对核函数的处理(规范化或中心化)会改变特征空间,使得度量结果失真。

关键词: 核函数,核函数度量,统计检验

Abstract: This paper explored the research on evaluating kernel function by using statistical testing.By employing kernel,normalized kernel,centered kernel and kernel distance as geometric measure among samples in feature space,and applying 7 statistical testing methods such as t-test and f-test,this paper evaluated the distributional difference between the geometric measures among samples from same classes and the geometric measure among samples from different classes.The experimental results of kernel selection on 11 UCI datasets show that the kernel evaluation measures based on statistical testing reach or exceed the performance of KTA and FSM,etc.And we found that the two types of data distribution differences are mainly reflected in the variance difference.Moreover,the formatting of kernel function such as normalization or centering can change the feature space,and make the evaluation distorted.

Key words: Kernel function,Kernel evaluation,Statistical testing

[1] Schlkopf B,Smola A.Learning with Kernels[M].MIT Press,Cambridge,Massachusetts,2002
[2] Lanckriet G R G,Cristianini N,Bartlett P,et al.Learning thekernel matrix with semi-definite programming[J].Journal of Machine Learning Research,2004,5:27-72
[3] Wang Cheng-qun,Chen Ji-ming,Hu Chong-hai,et al.Kernelmatrix learning with a general regularized risk functional criterion[J].Journal of Systems Engineering and Electronics,2010,21(1):72-80
[4] Girolami M,Rogers S.Hierarchic bayesian models for kernellearning[C]∥Proceedings of the 22nd international conference on machine learning.Bonn,Germany,2005:241-248
[5] Ong C S,Smola A J,Williamson R C.Learning the kernel with hyperkernels[J].Journal of Machine Learning Research,2005,6:1043-1071
[6] Yan Zhi-gang,Yang Yuan-xuan,Ding Yun-jing.An experimen-tal study of the hyper-parameters distribution region and its optimization method for support vector machine with gaussian kernel[J].International Journal of Signal Processing,Image Processing and Pattern Recognition,2013,6(5):437-446
[7] Cristianini N,Shawe-Taylor J,Elisseeff A,et al.On kernel-target alignment[J].Advances in Neural Information Processing Systems,2001,14:367-373
[8] Baram Y.Learning by kernel polarization[J].Neural Computation,2005,17:1264-1275
[9] Wang Ting-hua,Tian Sheng-feng,Huang Hou-kuan,et al.Learning by local kernel polarization[J].Neurocomputing,2009,72:3077-3084
[10] Nguyen C H,Ho Tu-bao.An efficient kernel matrix evaluation measure[J].Pattern Recognition,2008,41:3366-3372
[11] Wang Ting-hua,Zhao Dong-yan, Feng Yan-song.Two-stagemultiple kernel learning with multiclass kernel polarization[J].Knowledge-Based Systems,2013,48:10-16
[12] Zhong Shang-ping,Chen Da-ya,Xu Qiao-fen,et al.Optimizingthe gaussian kernel function with the formulated kernel target alignment criterion for two-class pattern classification[J].Pattern Recognition,2013,46:2025-2054
[13] Wang Lei,Xue Ping,Chan K L.Two criteria for model selection in multiclass support vector machines[J].IEEE Transactions on Systems,Man,and Cybernetics,2008,38(6):1432-1448
[14] Gnen M,Alpaydin E.Multiple kernel learning algorithms[J].Journal of Machine Learning Research,2011,12:2211-2268
[15] Wang Lei.Feature selection with kernel class separability[J].IEEE Transactions on pattern analysis and machine intelligence,2008,20(9):1524-1546
[16] Mao K Z.Feature subset selection for support vector machines through discriminative function pruning analysis[J].IEEE Transactions on Systems,Man,and Cybernetics,2004,34(1):60-67
[17] Vapnik V.Statistical Learning Theory[M].New York:Wiley-Interscience,1998
[18] John S-T,Cristianini N.Kernel methods for pattern analysis[M].Cambridge university press,2004
[19] Meilǎ M.Data centering in feature space [C]∥Ninth International Workshop on Artificial Intelligence and Statistics.2003
[20] Schlkopf B.The kernel trick for distance[J].Advances in Neural Information Processing Systems,2001,13:301-307
[21] Burago D,Burago Y D,Ivanov S.A Course in Metric Geometry[M].American Mathematical Society,2001
[22] Lesot M-J,Rifqi M.Similarity measures for binary and numerical data:a survey[J].International Journal of Knowledge Engineering and Soft Data Paradigms,2009,1(1):63-84
[23] Lomax R G.An Introduction to Statistical Concepts(Second Edition)[M].Routledge Academic,2007
[24] Wang Ting-hua,Zhao Dong-yan, Tian Sheng-feng.An overview of kernel alignment and its applications[J].Artificial Intelligence Review,2012(11)
[25] Chang Chih-Chung,Lin Chih-Jen.LIBSVM:a library for support vector machines[J].ACM Transactions on Intelligent Systems and Technology,2011,2(27):1-27

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!