Incomplete Multiview Clustering Based on Consensus Information.

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• Multiview clustering (MVC) integrates data from diverse sources for comprehensive analysis.
• Incomplete multiview clustering (IMVC) addresses challenges with missing data across views.
• Existing IMVC methods struggle with sample-wise differences and scalability due to matrix operations.

Purpose of the Study:

• To develop a novel IMVC method that overcomes limitations of existing approaches.
• To enhance clustering performance and computational efficiency for large-scale datasets.
• To propose a scalable and effective solution for incomplete multiview clustering.

Main Methods:

• A new multiview clustering with consistent information (IMVC-CI) approach is proposed.
• The method extracts consensus structural information from sample points across views.
• It restores missing information within each view independently, avoiding large matrix computations.

Main Results:

• The proposed IMVC-CI method demonstrates superior clustering performance compared to state-of-the-art methods.
• Significant improvements in computational efficiency were observed, especially for large datasets.
• The algorithm effectively handles incomplete multiview data by leveraging sample-wise consistency.

Conclusions:

• The novel IMVC-CI method offers an efficient and scalable solution for incomplete multiview clustering.
• By focusing on sample points and enabling parallel computation, it overcomes key limitations of prior work.
• The approach provides a robust framework for analyzing complex, multi-sourced datasets.