Anisotropic Graph Convolutional Network for Semi-Supervised Learning

Graph convolutional networks learn effective node embeddings that have proven to be useful in achieving high-accuracy prediction results semi-supervised learning tasks, such as classification. However, these suffer from the issue of over-smoothing and shrinking effect graph due large part fact they diffuse features across edges using a linear Laplacian flow. This limitation is especially problematic for task classification, where goal predict label associated with node. To address this issue, we propose an anisotropic network classification by introducing nonlinear function captures informative nodes, while preventing oversmoothing. The proposed framework largely motivated good performance diffusion image geometry processing, learns representations based on local structure features. effectiveness our approach demonstrated three citation two datasets, better or comparable accuracy compared standard baseline methods.