.png){kind=icon}
.png){kind=icon}
Return
Robust graph-regularized discriminative nonnegative matrix factorization for image clustering
.png){kind=icon}
PRE
AI
H
X
DOI：10.1007/s10489-026-07349-0.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Non-negative Matrix Factorization (NMF), as an effective tool for dimensionality reduction and clustering of high-dimensional data, is widely applied in fields such as computer vision, information retrieval, and pattern recognition. However, existing semi-supervised NMF methods, while achieving some improvements in clustering performance by leveraging supervisory information, remain limited in fully exploiting the intrinsic structure of data and resisting noise interference. To address these challenges, we propose a novel method, Robust Graph-Regularized Discriminative Nonnegative Matrix Factorization (RGD-NMF). Building upon Element Difference Discriminative NMF (EDDNMF), RGD-NMF integrates graph regularization and an $${L}_{\text{2,1}}$$ -norm loss function to establish a semi-supervised clustering framework that concurrently enhances discriminative power, preserves geometric structure, and ensures robustness against noise. Specifically, the model introduces an element difference discriminant term to increase inter-class separability, employs graph regularization to maintain the local manifold structure of the data, and adopts the $${L}_{\text{2,1}}$$ -norm for robust reconstruction, thereby achieving synergistic optimization of supervisory information, geometric structure, and noise resilience. Clustering experiments on multiple real-world datasets demonstrate that the proposed RGD-NMF method outperforms existing approaches in both clustering accuracy and robustness, confirming its effectiveness.
Keywords:
Nonnegative matrix factorization
Element difference discrimination
Semi-supervised Learning
Graph regularization
Robustness
Journal
IF:
3.5
Papers:
7.5K
Citations:
1.7W
Organization
Cited Papers