Abstract: Graph convolutional neural networks have demonstrated promising solutions for processing non-Euclidean data in tasks such as node classification. While existing graph convolution models aim ...