A low-rank non-convex norm method for multiview graph clustering

Abstract

This study addresses the challenge of multiview clustering by integrating information from multiple data sources to improve clustering accuracy. We propose CGMVC-NC, a novel Consensus Graph-Based Multi-View Clustering method Using Low-Rank Non-Convex Norm, which effectively captures correlations across views. Unlike traditional methods, CGMVC-NC introduces a non-convex low-rank tensor norm to enhance the representation of shared structures while reducing noise and redundancy. By constructing a consensus graph that preserves essential multiview relationships, our approach ensures more reliable clustering results. Extensive experiments on benchmark datasets confirm its superiority over existing techniques, demonstrating improved clustering performance and robustness in handling complex multiview data.