Semi-supervised Locality Preserving Projections with Compactness Enhancement

As the development of digital technique, dimensionality reduction becomes an indispensable measure to those tasks such as machine learning, pattern recognition and so on. Locality Preserving Projections (LPP) is a popular unsupervised dimensionality reduction technique in pattern recognition and machine learning, which finds the best projections by solving a variant problem that optimally preserves local neighborhood information of the data set. And it is a linear approximation of the nonlinear Laplacian Eigenmap. However, it may lose important information for it does not use any label information. In order to overcome this problem, we propose a method called Semi-supervised Locality Preserving Projections with Compactness Enhancement (SLPP), which can utilize both labeled and unlabeled samples simultaneously, and enhance the compactness of labeled samples. Experimental results on real world data sets confirm its effectiveness.