One-Step Late Fusion Multi-view Clustering with Compressed Subspace

TL;DR

This work tackles bottlenecks in late-fusion multi-view clustering by integrating consensus partition alignment with fused partition learning in a single framework. The proposed One-Step Late Fusion Multi-view Clustering with Compressed Subspace (OS-LFMVC-CS) leverages a shared consensus subspace and a compressed representation to directly learn discrete cluster labels, avoiding post-hoc spectral or k-means steps. A six-step iterative optimization updates views, fusion weights, the compressed subspace, and the discrete labels with guaranteed convergence and linear per-iteration complexity. Empirical results on five benchmark datasets show strong, scalable performance and efficient computation, validating the method's effectiveness for large-scale multi-view clustering tasks.

Abstract

Late fusion multi-view clustering (LFMVC) has become a rapidly growing class of methods in the multi-view clustering (MVC) field, owing to its excellent computational speed and clustering performance. One bottleneck faced by existing late fusion methods is that they are usually aligned to the average kernel function, which makes the clustering performance highly dependent on the quality of datasets. Another problem is that they require subsequent k-means clustering after obtaining the consensus partition matrix to get the final discrete labels, and the resulting separation of the label learning and cluster structure optimization processes limits the integrity of these models. To address the above issues, we propose an integrated framework named One-Step Late Fusion Multi-view Clustering with Compressed Subspace (OS-LFMVC-CS). Specifically, we use the consensus subspace to align the partition matrix while optimizing the partition fusion, and utilize the fused partition matrix to guide the learning of discrete labels. A six-step iterative optimization approach with verified convergence is proposed. Sufficient experiments on multiple datasets validate the effectiveness and efficiency of our proposed method.

Figures (2)

• Figure 1: Experimental results.
• Figure 2: Time comparison.