Multi-View Subspace-Enhanced Clustering Via Tensor-Nuclear Norm and Discriminative Graph

Multi-view data offer richer features than single-view data but are often noisy, posing challenges. To address these issues, we propose a novel multi-view clustering model that integrates weighted tensor-nuclear norm regularization and discriminative graph learning. Our approach decomposes multi-view features into self-representative and noise components. It uses a weighted tensor-nuclear norm to capture higher-order correlations among samples, adaptively assigning weights based on clustering contributions. The noise term is constrained by the L<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>-norm to enhance robustness. Additionally, a discriminative graph learning framework is designed to model inter-class commonality and intra-class distinctiveness jointly. Extensive experiments show that our method outperforms state-of-the-art techniques, demonstrating its effectiveness.