Consistent Subspace-Based Multi-View Multi-Label

This study proposes an innovative multi-view multi-label learning algorithm by constructing a commonality subspace, which projects multi-view data into a shared space to better integrate information from different views. Label prediction in this commonality subspace not only enhances the synergistic effects between views but also effectively addresses the heterogeneity among views. Additionally, the method incorporates regularization strategies to further improve the model's generalization capability and reduce the risk of overfitting. Experimental results demonstrate that, compared to traditional methods, this approach significantly improves multi-label classification performance across multiple benchmark datasets, validating its effectiveness and robustness in multi-view multi-label learning tasks. Furthermore, parameter sensitivity analysis and ablation studies show that the method maintains good stability and efficiency under various parameter settings, indicating its broad application prospects.