Multi-View Maximum Entropy Clustering by Jointly Leveraging Inter-View Collaborations and Intra-View-Weighted...

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Multi-view clustering is crucial for heterogeneous data but existing methods often overlook inter-view collaborations or attribute importance.
• Current approaches may treat attributes within views as equally important, limiting their effectiveness.

Purpose of the Study:

• To propose a novel collaborative multi-view clustering model, VC-AW-MEC, addressing limitations of existing methods.
• To enhance clustering by incorporating inter-view collaborative learning (IEVCL) and intra-view attribute weighting (IAVWA) into Maximum Entropy Clustering (MEC).

Main Methods:

• Developed two criteria: inter-view collaborative learning (IEVCL) and intra-view attribute weighting (IAVWA).
• Integrated IEVCL and IAVWA into the Maximum Entropy Clustering (MEC) framework to create the VC-AW-MEC model and algorithm.
• Validated the approach on synthetic and real-world heterogeneous multi-view datasets.

Main Results:

• The proposed VC-AW-MEC effectively handles diverse multi-view data scenarios.
• IEVCL successfully seeks consensus across views, while IAVWA adaptively weighs attribute importance within each view.
• VC-AW-MEC demonstrated superior clustering effectiveness and stability compared to state-of-the-art methods.

Conclusions:

• The novel VC-AW-MEC model offers a significant advancement in multi-view clustering for heterogeneous data.
• Jointly leveraging IEVCL and IAVWA provides a robust approach to improve clustering performance.
• The method's effectiveness is confirmed across various data types and complexities.