Sparse Locality Preserving Embedding

Linear dimensionality reduction algorithms, such as principal component analysis, linear discriminant analysis and locality preserving projections, have attracted much attention in many fields. However, the embedding results obtained by those algorithms are linear combination of all the original features, which is difficult to be interpreted psychologically and physiologically. This paper proposes a novel technique, called sparse locality preserving embedding, which performs in the lasso regression framework that dimensionality reduction, feature selection and classification are merged into one analysis. Additionally, the algorithm can be performed both in supervised and unsupervised tasks. Experimental results show that our methods are effective and demonstrate much higher performance.