Fast Multiview Clustering With Spectral Embedding.

Area of Science:

• Computer Science
• Machine Learning
• Data Mining

Background:

• Spectral clustering is effective for unsupervised learning but struggles with large-scale and high-dimensional data.
• Existing methods face computational complexity issues, especially with multi-view datasets.

Purpose of the Study:

• To propose a fast multi-view clustering algorithm with spectral embedding (FMCSE).
• To enhance the efficiency of spectral embedding and spectral analysis stages in multi-view spectral clustering.
• To enable direct acquisition of sample categories post-optimization, eliminating the need for k-means.

Main Methods:

• Developed FMCSE algorithm to accelerate spectral embedding and analysis.
• Implemented a fast optimization strategy by decoupling the problem into smaller sub-problems.
• Avoided traditional k-means post-processing for category acquisition.

Main Results:

• FMCSE significantly improves clustering efficiency for large-scale and high-dimensional datasets.
• The algorithm maintains comparable or superior clustering effectiveness against state-of-the-art baselines.
• Experiments on diverse real-world datasets validate the proposed method's performance.

Conclusions:

• FMCSE offers an efficient and effective solution for multi-view spectral clustering on large datasets.
• The decoupled optimization strategy contributes to the algorithm's speed and scalability.
• This work advances unsupervised learning for complex, large-scale data analysis.