Two Decades of Multi-View Clustering: Taxonomy, Application, and Challenge

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Multi-view clustering (MVC) leverages complementary information from multiple data sources.
• Significant advancements have been made in MVC over the last two decades.
• A comprehensive summary of existing MVC methods and future challenges is lacking.

Purpose of the Study:

• To provide a thorough review of existing multi-view clustering methods.
• To categorize MVC methods based on techniques, fusion strategies, and scenarios.
• To identify and discuss future research directions and challenges in MVC.

Main Methods:

• Categorization of MVC methods into seven techniques, four fusion strategies, and five scenarios.
• Collection and analysis of commonly used datasets for MVC.
• Performance analysis of typical MVC methods on benchmark datasets.

Main Results:

• A structured overview of current MVC methodologies.
• Identification of six diverse application areas, including computer vision, information retrieval, and bioinformatics.
• Seven promising future research directions are proposed for the field.

Conclusions:

• The survey offers a systematic review and taxonomy of multi-view clustering.
• It provides insights into current applications and performance benchmarks.
• The identified future directions aim to guide subsequent research in multi-view clustering.