I. Introduction
Subpace clustering (SC) [1], [2], [3], [4], [5], [6], [7], especially self-representation subspace clustering methods [8], [9], [10], [11], [12], [13], has gained extensive attention for real-world applications [14], such as image clustering, motion segmentation, and object tracking [15], [16], [17]. For a series of recent works [18], [19], [20], [21], [22], [23], [24], the main idea is assuming that data points can be self-represented with an affinity matrix, which is subsequently used for constructing the similarity graph for later clustering.