One-Step Multiview Fuzzy Clustering With Collaborative Learning Between Common and Specific Hidden Space Information.

Area of Science:

• Data Science
• Machine Learning
• Artificial Intelligence

Background:

• Multiview data is prevalent in real-world applications, necessitating effective mining techniques.
• Existing multiview clustering algorithms often focus on shared hidden spaces, facing challenges in capturing both shared and specific information and optimizing for clustering tasks.

Purpose of the Study:

• To propose a novel one-step multiview fuzzy clustering (OMFC-CS) method.
• To address challenges in learning hidden spaces that contain both shared and specific information.
• To design an efficient mechanism for integrating hidden space learning with fuzzy partitioning for enhanced clustering.

Main Methods:

• Developed a matrix factorization-based mechanism for simultaneous extraction of common and specific information from multiview data.
• Designed a one-step learning framework integrating common/specific space learning and fuzzy partition learning through alternate, mutually beneficial processes.
• Incorporated Shannon entropy for optimal view weight assignment during clustering.

Main Results:

• The proposed OMFC-CS method effectively extracts both shared and specific information from multiview data.
• The integrated one-step learning framework enhances the suitability of learned hidden spaces for clustering.
• Experimental results on benchmark datasets show OMFC-CS outperforms existing multiview clustering methods.

Conclusions:

• The OMFC-CS method offers a significant advancement in multiview clustering by collaboratively learning common and specific information.
• The one-step integrated framework provides a more effective approach to hidden space learning and fuzzy clustering.
• OMFC-CS demonstrates superior performance, highlighting its potential for real-world multiview data analysis.