Robust Subspace Clustering via Local Weighted sparse and Global Prior information
Journal of Decision Making and Healthcare, Volume 3, Issue 2, August 2026, Pages: 62–77
JIAQI ZHANG
Chongqing Jiaotong University, Chongqing 400074, P.R. China
RENLI LIANG
Chongqing Jiaotong University, Chongqing 400074, P.R. China
ZAIYUN PENG
School of Mathematics, Yunnan Normal University, Kunming 650092, P.R. China
Abstract
The task of subspace clustering is to divide high-dimensional data into several low dimensional subspaces. Most existing methods ignore the local and global prior information of the data. In response to such issues, we propose a subspace clustering model based on local weighted sparsity and global prior information. By utilizing the distance information of the data, the guidance coefficient matrix \(Z\) is represented with as few local points as possible, which can enhance the discriminative ability of our model. By introducing a matrix \(D\) to describe different prior information of data. In addition, considering that real-world data may contain multiple types of noise, we consider sparse noise and Gaussian noise priors to enhance the robustness of our model. Then, we have introduced an efficient splitting algorithm based on the augmented Lagrangian framework for the model to ensure its convergence. Finally, experiments on synthetic data, four real-world datasets, temporal action datasets, and manifold datasets demonstrate that our method outperforms existing methods in clustering accuracy, computational efficiency, and noise robustness.
Cite this Article as
Jiaqi Zhang, Renli Liang and Zai Yun Peng, Robust Subspace Clustering via Local Weighted sparse and Global Prior information, Journal of Decision Making and Healthcare, 3(2), 62–77, 2026