Tensor Learning Meets Dynamic Anchor Learning: From Complete to Incomplete Multiview Clustering.

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Multiview clustering (MVC) effectively reveals intrinsic data structures.
• Existing MVC methods are limited to either complete or incomplete datasets.
• A unified framework for both scenarios is lacking.

Purpose of the Study:

• To propose a unified framework for multiview clustering (MVC) that handles both complete and incomplete data simultaneously.
• To develop a scalable and efficient MVC method with approximately linear complexity.
• To leverage tensor learning and dynamic anchor learning for enhanced clustering performance.

Main Methods:

• Developed TDASC (Tensor and Dynamic Anchor learning for Scalable Clustering).
• Integrated tensor learning to capture inter-view low-rankness and high-order correlations.
• Employed dynamic anchor learning for intra-view low-rankness and efficient graph construction.

Main Results:

• TDASC achieves approximately linear complexity for scalable clustering.
• The framework effectively models high-order correlations across multiple views.
• Experiments demonstrate TDASC's superior effectiveness and efficiency on complete and incomplete datasets compared to state-of-the-art methods.

Conclusions:

• TDASC offers a unified and efficient solution for both complete and incomplete multiview clustering.
• The integration of tensor and anchor learning significantly enhances clustering accuracy and scalability.
• TDASC represents a significant advancement in multiview clustering research.