Several orthogonal feature extraction algorithms based on local preserving projection have recently been proposed. However, these methods still are linear techniques in nature. In this paper, we present nonlinear feature extraction method called Kernel Orthogonal Neighborhood Preserving Discriminant Analysis (KONPDA). A major advantage of the proposed method is that it is regarded every column of the kernel matrix as a corresponding sample. Then running KONPDA in kernel matrix, nonlinear features can be extracted. Experimental results on ORL database indicate that the proposed KONPDA method achieves higher recognition rate than the ONPDA method and other kernel-based learning algorithms.
Unsupervised Nonlinear Feature Extraction Method And Its Effects On Target Detection In High-Dimensional Data
