The Research of Face Recognition Based on Kernel Function

Abstract:

The way of kernel function has been widely applied in machine learning field, such as artificial neural network and support vector machine, for avoiding dimensional disaster of feature space and improving performance of learning machine effectively. The selection of kernel function and construction of new kernel are the core problems, which have a direct relation with the performance of classification, and the research of this field is not enough. In this paper support vector machine (SVM) was used as an example, and the performance of common kernel functions was evaluated through observing and computing the features of kernel matrix. Base on this, a new mixed kernel function was gotten by optimization of kernel functions, and the experimental data proved that the performance of SVM was improved by the mixed kernel function. If the weighting coefficient was selected properly, the correct rate could even reach to 100%. What’s more, not only a method to construct a new learning machine was given, but also a reference for selecting kernel function was given.

• Periodical: Applied Mechanics and Materials (Volumes 321-324)
• Pages: 1181-1185
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.321-324.1181
• Online since: June 2013
• Authors: Yong Jun Zhu, Wen Bo Liu, Feng Yu Jin, Yin Wu
• Keywords: Kernel Function, Kernel Matrix, Neural Network (NN), Support Vector Machine (SVM)

References

[1] V. N. Vapnik, Xuegong Zhang: Basis on statistical learning theory. Tsinghua University Press, Beijing(2000).

[2] Shawe-Taylor, N Cristianini: Kernel methods for pattern analysis. Cambridge University Press, England(2004).

[3] S. Wu, S. Amari: Conformal transformation of kernel functions: a data-dependent way to improve support vector machine classifiers. Neural Processing Letters(S1370-4621),Vol.15 (2002), pp.59-67.

[4] O Chapelle, V N Vapnik et al: Choosing multiple parameters for support vector machines. Machine Learning, Vol.46(2002),pp.131-159.

[5] Jingcheng Liu, Shunpeng Zeng: Intelligent evaluation model for cementing quality based on GA-SVM and application. Applied Mechanics and Materials, Vol.121-126(2012), pp.2730-2734.

DOI: 10.4028/www.scientific.net/amm.121-126.2730

[6] N. Cristianini, J. Shawe-Taylor, J. Kandola, et al: On kernel target alignment. Neural Information Processing Systems. MA: MIT Press, Cambridge(2002),pp.367-373.

DOI: 10.7551/mitpress/1120.003.0052

[7] Xiangdong Liu, Bing Luo，Zhaoqian Cheng: Research of optimal model choice on SVM. Research and development of computer, Vol.42 (4) (2005),p.576～581.

[8] Feng Zhang, Liang Tao, Yan Sun: SVM based on mixed kernel function and application. Technology and development of computer, Vol.16(2)(2006),pp.176-178.

[9] Suykens J A K: Nonlinear modeling and support vector machine. IEEE Instrumentation and Measurement Technology Conference Budapest. Hungary(2001),pp.21-23.

[10] Solla S A, Leen T K, Muller K R: Advances in Neural Information Processing System. MIT Press, Cambridge(2000),pp.230-237.