Dual Consensus Anchor Learning for Fast Multi-View Clustering.

Area of Science:

• Machine Learning
• Data Mining
• Computer Science

Background:

• Multi-view clustering methods integrate graph structures for improved performance.
• Anchor-based methods reduce computation costs for large datasets.
• Existing methods lack guaranteed cluster structure correspondence between anchor graphs and partitions.

Purpose of the Study:

• To propose a novel Dual consensus Anchor Learning for Fast multi-view clustering (DALF) method.
• To ensure cluster structure correspondence between anchor graphs and partitions in large-scale multi-view datasets.
• To discover anchor graphs depicting shared cluster assignments across views under orthogonal constraints.

Main Methods:

• DALF jointly learns anchors, constructs anchor graphs, and performs partitions within a unified framework.
• Employs rank constraint on the Laplacian graph and orthogonal constraint on centroid representation.
• Introduces orthogonal constraint in anchor graph factorization for direct cluster assignment construction.

Main Results:

• DALF guarantees cluster structure correspondence between anchor graph and partition for large-scale multi-view data.
• The method simultaneously optimizes for cluster structure in both anchor graph and partition.
• Extensive experiments validate the effectiveness and efficiency of DALF compared to existing methods.

Conclusions:

• DALF offers an effective and efficient solution for large-scale multi-view clustering.
• The proposed method ensures robust cluster structure representation across different views.
• DALF advances the field by addressing limitations in existing anchor-based multi-view clustering techniques.