High-Order Correlation Preserved Incomplete Multi-View Subspace Clustering.

Area of Science:

• Machine Learning
• Data Mining
• Computer Vision

Background:

• Incomplete multi-view clustering methods often overlook high-order correlations.
• Existing approaches focus on pairwise sample and view correlations, limiting performance.

Purpose of the Study:

• To propose a novel method, High-Order Correlation Preserved Incomplete Multi-View Subspace Clustering (HCP-IMSC), for improved clustering of incomplete multi-view data.
• To effectively recover missing views and preserve the subspace structure of incomplete multi-view data.

Main Methods:

• Utilizing tensor factorization to preserve high-order correlations in multiple incomplete views.
• Constructing a unified affinity matrix through self-weighted fusion of view-specific matrices.
• Employing hypergraph construction and hyper-Laplacian regularization for data reconstruction.

Main Results:

• The HCP-IMSC method demonstrates superior performance on benchmark datasets compared to existing methods.
• Experimental results validate the effectiveness of preserving high-order correlations for clustering incomplete multi-view data.

Conclusions:

• The proposed HCP-IMSC method offers a significant advancement in incomplete multi-view clustering.
• Integrating tensor factorization and hypergraph regularization provides a robust framework for handling missing data and complex correlations.