Title :
A modified PCA based on the minimum error entropy
Author :
Guo, Zhenhua ; Yue, Hong ; Wang, Hong
Author_Institution :
Sch. of Mech. Sci. & Eng., Huazhong Univ. of Sci. & Technol., Hubei, China
fDate :
June 30 2004-July 2 2004
Abstract :
Conventional principal component analysis (PCA) minimizes the total error variance, which may be inappropriate for the non-Gaussian distribution systems. In this paper the entropy is proposed as a more general index for PCA model, and then a modified PCA with the optimization for the minimum error entropy via a genetic algorithm (GA) is addressed.
Keywords :
genetic algorithms; least mean squares methods; minimum entropy methods; principal component analysis; GA optimization; genetic algorithm; minimum error entropy; modified PCA; nonGaussian distribution systems; principal component analysis; total error variance;
Conference_Titel :
American Control Conference, 2004. Proceedings of the 2004
Conference_Location :
Boston, MA, USA
Print_ISBN :
0-7803-8335-4
