Multi-view clustering via global-view graph learning.

Multiview clustering aims to improve clustering performance by exploring multiple representations of data and has become an important research direction. Meanwhile, graph-based methods have been extensively studied and have shown promising performance in multiview clustering tasks. However, most existing graph-based multiview clustering methods rely on assigning appropriate weights to each view based on its importance, with the clustering results depending on these weight assignments. In this paper, we propose an a novel multiview spectral clustering framework with reduced computational complexity that captures complementary information across views by optimizing a global-view graph using adaptive weight learning. Additionally, in our method, once the Global-view Graph is obtained, cluster labels can be directly assigned to each data point without the need for any post-processing, such as the K-means required in standard spectral clustering. Our method not only improves clustering performance but also reduces computational resource consumption. Experimental results on real-world datasets demonstrate the effectiveness of our approach.