Correntropy-Based Multiview Subspace Clustering

Area of Science:

• Data Mining
• Pattern Recognition
• Machine Learning

Background:

• Multiview subspace clustering integrates data from multiple sources for enhanced pattern recognition.
• Challenges include efficiently learning representation matrices and leveraging inter-view information.

Purpose of the Study:

• To propose a novel correntropy-based multiview subspace clustering (CMVSC) method.
• To address challenges in learning from multiple data views for improved clustering.

Main Methods:

• Developed a CMVSC model with a two-part objective function.
• Utilized Frobenius norm for estimating subspace connections and correntropy-induced metric (CIM) for noise characterization and information fusion.
• Employed half-quadratic (HQ) and alternating direction method of multipliers (ADMM) for optimization.

Main Results:

• The proposed CMVSC method demonstrated superior performance on six real-world multiview datasets.
• Outperformed several state-of-the-art multiview subspace clustering techniques.

Conclusions:

• CMVSC effectively addresses the limitations of single-view clustering.
• The method shows significant potential for applications in data mining and pattern recognition.