Entropy-based active sparse subspace clustering

Sparse Subspace Clustering (SSC) is widely used in data mining and machine learning. Some studies have been developed to add pairwise constraints as side information to improve the clustering results. However, most of these algorithms are “passive” in the sense that the side information is provided beforehand. In this paper, we propose a novel extension for SSC with active learning framework, in which we aim to select the most informative pairwise constraints to guide the SSC for accurate clustering results. Specifically, in the first step, an entropy-based query strategy is proposed to select the most uncertain pairwise constraints. Next, constrained sparse subspace clustering algorithms are followed to integrate the selected pairwise constraints and obtain the final clustering results. Two steps are effectively performed in an iterative manner until satisfactory results are achieved. Experimental results on two face datasets clustering well demonstrate the effectiveness of the proposed method.