Efficient Discrete Multi-View Clustering with Tensorized Label Learning
Qin Li, Geng Yang, Bob Zhang, Ming Yang
Source abstract
Anchor graphs are widely recognized for their capability to reduce the computational complexity of clustering. When regression techniques are applied to anchor graphs, existing regression-based methods mainly focus on directly learning sample labels, while the label distribution induced on anchors through sample–anchor relationships remains insufficiently explored. We propose Efficient Discrete Multi-View Clustering with Tensorized Label Learning (EDMVC_TLL) to address this limitation. Specifically, we fully leverage the probabilistic properties of anchor graphs to establish the relationship between sample labels and anchor labels, utilizing anchor labels to guide the learning of sample labels. Furthermore, we introduce an innovative regularization term to balance anchor labels, avoiding trivial solutions and enhancing sample labels’ quality. The tensor Schatten p-norm explores complementary information embedded in different views. Additionally, we adopt an adaptive anchor graph, integrating it into the subsequent optimization process to improve the quality of the anchor graph. An efficient optimization algorithm is developed, and extensive experiments on six benchmark datasets demonstrate that EDMVC_TLL consistently achieves the best performance in terms of ACC, NMI, and Purity among the compared methods.
Evidence graph
No public relationships recorded yet.
Integrity note: This page is a factual metadata record created by deterministic ingestion. It is not a claim that the work moves a mathematical frontier or has been independently verified.