Local-Structure Adaptive Sparse Subspace Learning: An Iterative Approach to Unsupervised Feature Selection

Subspace learning is becoming more and more popular thanks to its capabilities of good interpretation. However, existing approaches do not adapt both local structure and self reconstruction very well. We propose local-structure adaptive sparse subspace learning (ASSL) model for unsupervised feature selection. In this paper, we formulate the feature selection process as a subspace learning problem and incorporate a regularization term to preserve the local structure of the data. Furthermore, we develop a greedy algorithm to establish the basic model and an iterative strategy based on an accelerated block coordinate descent is used to solve the local-structure ASSL problem. We also provide the global convergence analysis of the proposed ASSL algorithm. Extensive experiments are conducted on real-world datasets to show the superiority of the proposed approach over several state-of-the-art unsupervised feature selection approaches.