Multi-View Representation Learning with Manifold Smoothness

Multi-view representation learning attempts to learn a from multiple views and most existing methods are unsupervised. However, learned only unlabeled data may not be discriminative enough for further applications (e.g., clustering classification). For this reason, semi-supervised which could use along with the labeled multi-view need developed. Manifold information plays an important role in learning, but it has been considered learning. In paper, we introduce manifold smoothness into propose MvDGAT learns intrinsic simultaneously graph attention network. Experiments conducted on real-world datasets reveal that our can achieve better performance than state-of-the-art methods.