Unsupervised feature selection via graph matrix learning and the low-dimensional space learning for classification

摘要

Unsupervised feature selection is a powerful tool to select a subset of features for effective representation of high-dimensional data. In this paper, we proposes a novel unsupervised feature selection method via the graph matrix learning and the low-dimensional space learning to obtain their individually optimized result. Furthermore, the global and local correlation of features have been taken into consideration through the low-rank constraint and the feature-level representation property on the graph matrix. Experimental analysis on 15 benchmark datasets verified that our proposed method outperformed the state-of-the-art feature selection methods in terms of classification performance.