Semi-Supervised Multi-View Clustering with Weighted Anchor Graph Embedding.

Area of Science:

• Machine Learning
• Data Mining
• Artificial Intelligence

Background:

• Multi-view clustering excels with complete data but struggles with real-world incomplete datasets and limited labels.
• Existing models often fail to effectively handle incomplete multi-view data or leverage labeled samples efficiently.
• This limits the practical application of current multi-view clustering techniques.

Purpose of the Study:

• To propose a novel framework, Semi-supervised Multi-View Clustering with Weighted Anchor Graph Embedding (SMVC_WAGE), for high-quality clustering on incomplete multi-view data.
• To address the limitations of existing methods in handling missing samples and efficiently utilizing labeled data.
• To improve both clustering performance and computational efficiency.

Main Methods:

• Introduced a simple and effective anchor strategy to bridge samples and capture nonlinear relations across views.
• Developed a parameter-free graph fusion mechanism to construct a global fused graph from view-wise graphs.
• Designed the SMVC_WAGE framework to be conceptually simple yet efficient for practical application.

Main Results:

• The proposed SMVC_WAGE framework effectively handles both complete and incomplete multi-view clustering scenarios.
• Experimental results demonstrate superior clustering ability compared to state-of-the-art competitors.
• The algorithm also shows a significant reduction in time cost, indicating improved efficiency.

Conclusions:

• SMVC_WAGE offers a robust and efficient solution for multi-view clustering, particularly in challenging incomplete data settings.
• The method successfully integrates semi-supervised learning with an anchor-based graph embedding approach.
• The framework's ability to handle missing data and reduce computational complexity makes it suitable for practical applications.