Multiview Clustering via Unified and View-Specific Embeddings Learning.

Area of Science:

• Machine Learning
• Data Science
• Artificial Intelligence

Background:

• Multiview clustering leverages multiple feature sets to enhance clustering accuracy.
• Subspace-based methods learn unified embeddings but often neglect data similarity rankings.
• Existing methods struggle with incomplete multiview data, significantly degrading performance.

Purpose of the Study:

• To propose a novel multiview clustering approach inspired by knowledge graph embeddings.
• To integrate unified and view-specific embedding learning by treating views as relations.
• To develop a method robust to incomplete multiview data.

Main Methods:

• Representing different data views as relations within a knowledge graph.
• Employing embedding techniques from natural-language processing for unified and view-specific learning.
• Extending the framework to handle incomplete multiview datasets.

Main Results:

• The proposed method achieves superior performance compared to state-of-the-art clustering algorithms.
• Demonstrated effectiveness in utilizing incomplete feature sets for improved clustering.
• Successful application of knowledge graph embedding principles to multiview clustering.

Conclusions:

• The novel knowledge graph-inspired multiview clustering method offers significant improvements.
• The approach effectively addresses the challenge of incomplete multiview data.
• This work provides a promising direction for future research in multiview clustering.