A supervised subspace learning algorithm: Supervised Neighborhood Preserving Embedding

Abstract

Neighborhood Preserving Embedding (NPE) is an unsupervised manifold learning algorithm with subspace learning characteristic. In fact, NPE is a linear approximation to Locally Linear Embedding (LLE). So it can provide an unsupervised subspace learning technique. In this paper, we proposed a new Supervised Neighborhood Preserving Embedding (SNPE) algorithm which can use the label or category information of training samples to better describe the intrinsic structure of original data in low-dimensional space. Furthermore, when a new unknown data needs to be processed, SNPE, as a supervised subspace learning technique, may be conducted in the original high-dimensional space. Several experiments on USPS digit database demonstrate the effectiveness of our algorithm. © Springer-Verlag Berlin Heidelberg 2007.