Unsupervised feature selection by combining subspace learning with feature self-representation

摘要: Abstract So far, most of existing feature selection methods have two defects: 1) These are more or less heavy workload, 2) Their effect is not very good enough. To solve the above issues, a novel unsupervised algorithm proposed. Specifically, proposed method uses property data to construct self-representation coefficient matrix, and utilizes sparse representation find structure embeds hypergraph Laplacian regularization term make up insignificance ordinary graph in multiple relations. The Linear Discriminant Analysis (LDA) used further adjust result selection. Finally, low rank constraint capture global data. Experimental results on real datasets showed that outperformed state-of-the-art methods.