Linearity-Aware Subspace Clustering

Obtaining a good similarity matrix is extremely important in subspace clustering. Current state-of-the-art methods learn the through self-expressive strategy. However, these directly adopt original samples as set of basis to represent itself linearly. It difficult accurately describe linear relation between real-world applications, and thus hard find an ideal matrix. To better samples, we present clustering model, Linearity-Aware Subspace Clustering (LASC), which can consciously by employing linearity-aware metric. This new method that combines metric learning into joint framework. In our first utilize strategy obtain initial structure discover low-dimensional representation data. Subsequently, use proposed intrinsic with on obtained subspace. Based such learned matrix, inter-cluster distance becomes larger than intra-cluster distances, successfully obtaining cluster result. addition, enrich more consistent knowledge, collaborative for learning. Moreover, provide detailed mathematical analysis show properly characterize correlation samples.