Hyper-Laplacian regularized multi-view subspace clustering with low-rank tensor constraint.

Area of Science:

• Machine Learning
• Data Mining
• Computer Vision

Background:

• Real-world data often exhibits complex structures across multiple views.
• Existing multi-view clustering methods may struggle with high-order correlations and nonlinear subspace structures.
• Preserving local geometric information is crucial for accurate clustering.

Purpose of the Study:

• To propose a novel multi-view subspace clustering method that leverages hyper-Laplacian regularization and low-rank tensor constraints.
• To effectively capture high-order correlations among different views.
• To preserve the local geometric structure of data residing in nonlinear subspaces.

Main Methods:

• A hyper-Laplacian regularized multiview subspace clustering with low-rank tensor constraint (HLR-MSCLRT) model is proposed.
• Subspace representation matrices from different views are stacked into a tensor to capture high-order correlations.
• A low-rank constraint is applied to the tensor to reduce redundancy.
• Hyper-Laplacian graph regularization is used to preserve local geometry in high-dimensional spaces.

Main Results:

• The HLR-MSCLRT model effectively captures high-order correlations among data from multiple views.
• The method successfully preserves the local geometric structure of data within nonlinear subspaces.
• Experimental results demonstrate that HLR-MSCLRT outperforms existing state-of-the-art multi-view clustering approaches.

Conclusions:

• The proposed HLR-MSCLRT method offers a robust framework for multi-view subspace clustering.
• The integration of hyper-Laplacian regularization and low-rank tensor constraints enhances clustering accuracy.
• HLR-MSCLRT shows significant potential for applications requiring advanced multi-view data analysis.