Fast Multiview Anchor-Graph Clustering.

Area of Science:

• Computer Science
• Data Science
• Machine Learning

Background:

• Graph-based methods face computational challenges in large-scale multiview clustering.
• Existing accelerated methods often approximate discrete problems, leading to potential effectiveness and efficiency loss.
• Anchor graph and indicator learning methods have shown success but still rely on approximation strategies.

Purpose of the Study:

• To develop a discrete multiview clustering model that avoids approximation and discretization issues.
• To enhance both the effectiveness and efficiency of large-scale multiview clustering.
• To propose a novel optimization strategy for direct discrete graph-cutting problem solving.

Main Methods:

• A discrete fast multiview anchor graph clustering (FMAGC) model is established.
• Anchor graphs are constructed for each view.
• A discrete multiview graph-cutting problem is solved directly using a fast coordinate descent-based optimization strategy with linear complexity.

Main Results:

• The proposed FMAGC model effectively addresses the limitations of approximation and discretization.
• The coordinate descent optimization strategy achieves linear complexity, enabling efficient computation.
• Experiments demonstrate improved clustering effectiveness and efficiency compared to state-of-the-art methods on various datasets.

Conclusions:

• FMAGC offers a superior approach to large-scale multiview clustering by directly handling discrete graph-cutting.
• The method provides a balance between effectiveness and computational efficiency.
• This work advances the field of multiview clustering by offering a practical and high-performing solution.