Labeled diffusion-constrained nonnegative matrix factorization for multiview clustering

doi:10.1016/j.engappai.2026.113977

Labeled diffusion-constrained nonnegative matrix factorization for multiview clustering

2026-02-03

0

PRE

AI

DOI：10.1016/j.engappai.2026.113977

Original

Original request for help

Save

Abstract

En 中文

Multi-view clustering has become a popular approach for integrating information from diverse data sources. However, making full use of limited label information remains a major challenge. To address this, we propose a novel labeled diffusion-constrained non-negative matrix factorization (LCNMF) method. Our method is designed to effectively utilize partial label information by introducing a label consistency mechanism. Specifically, we generate a label constraint matrix and embed it into the similarity matrix to enhance its accuracy, while also transforming global constraints into local ones to further improve performance. We developed an efficient multiplicative update algorithm to solve the optimization problem and proved its convergence. Extensive experiments on five benchmark datasets demonstrate that our proposed algorithm achieves a significant average performance improvement of 3.2% compared to state-of-the-art methods, confirming its effectiveness.

Labeled diffusion-constrained nonnegative matrix factorization for multiview clustering

Abstract

Journal

Organization