Multiview Clustering Based on Non-Negative Matrix Factorization and Pairwise Measurements.

Area of Science:

• Machine Learning
• Pattern Recognition
• Data Mining

Background:

• Multiview clustering is a significant area in machine learning.
• Non-negative Matrix Factorization (NMF) is a popular technique for multiview clustering.
• Existing NMF-based methods often overlook inter-view similarities.

Purpose of the Study:

• To propose a novel multiview clustering algorithm.
• To enhance NMF by incorporating pairwise co-regularization and manifold regularization.
• To address the limitations of existing methods by considering inter-view similarities.

Main Methods:

• Developed a multiview clustering algorithm integrating NMF with pairwise co-regularization and manifold regularization.
• Utilized pairwise co-regularization to capture inter-view similarities and create a compact data representation.
• Employed NMF for part-based representation and manifold regularization to preserve local geometric structures.

Main Results:

• The proposed algorithm effectively considers both intra-view and inter-view similarities.
• Achieved a more compact and part-based representation of multiview data.
• Demonstrated superior performance compared to state-of-the-art methods in experimental evaluations.
• Provided theoretical proof of the algorithm's convergence.

Conclusions:

• The novel algorithm offers an effective approach to multiview clustering.
• Integrating inter-view similarity enhances data representation and clustering accuracy.
• The proposed method represents a significant advancement in the field of multiview clustering.