Local sampling mean discriminant analysis with kernels
Guiyu Feng, H.-T. Xiao, Qiang Fu · Electronics Letters · 2012
To overcome the drawbacks of linear discriminant analysis, such as homogeneous samples with Gaussian distribution and the small number of available projection vectors, local sampling mean discriminant analysis (LSMDA) has been proposed recently. In this Letter, the kernel LSMDA is proposed to alleviate the loss of class discrimination after linear feature extraction. Experimental results on ten UCI datasets demonstrate the efficiency of the proposed method.