Regularized Instance Weighting Multiview Clustering via Late Fusion Alignment.

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• Multiview clustering is vital for data analysis across diverse fields.
• Existing late fusion multiview clustering (LFMVC) methods struggle with varying instance importance and noise sensitivity.
• Effective alignment and fusion of information from multiple data sources remain challenging.

Purpose of the Study:

• To propose a novel regularized instance weighting multiview clustering via late fusion alignment (R-IWLF-MVC).
• To enhance information integration by considering instance importance and mitigating noise influence.
• To improve the robustness and effectiveness of multiview clustering.

Main Methods:

• Developed a regularized instance weighting approach (R-IWLF-MVC) for multiview clustering.
• Assigned importance attributes to samples to focus learning on key instances and reduce outlier impact.
• Employed late fusion alignment with a novel regularization term incorporating prior knowledge.
• Designed a three-step alternating optimization strategy with proven convergence.

Main Results:

• The proposed R-IWLF-MVC method effectively addresses limitations of existing LFMVC approaches.
• Instance weighting improves information integration and reduces sensitivity to noise and outliers.
• Evaluations on multiple real-world datasets demonstrate superior performance compared to state-of-the-art methods.

Conclusions:

• R-IWLF-MVC offers a robust and effective solution for multiview clustering.
• The method's ability to handle instance importance and noise makes it suitable for complex data.
• This work advances the field of multiview clustering with practical implications for data analysis.