Enhance Information Propagation for Graph Neural Network by Heterogeneous Aggregations

Success of deep learning in computer vision and natural language processing has recently boosted flood of research on applying neural networks to graph data (Wu et al., 2020). Graph is a simple yet versatile data structure jointly described by sets of nodes and edges. Aside from image and text data we're familiar, lots of real world data are better described as graph and thus processed by graph neural networks, such as social networks (Fan et al., 2019), financial fraud detection (Wang et al., 2020), knowledge graph (Zhang et al., 2020), biology interaction network (Higham et al., 2008), small molecule in drug discovery (Hu et al., 2019), to name a few. Since the seminal works (Kipf and Welling, 2016; Hamilton et al., 2017), tens of different graph neural network variants have been proposed, emphasizing different graph properties and design options. GNN research routes can be roughly divided into two categories: spectral based and spatial based. Spectral based GNNs try to approximate CNN's convolution by defining Fourier transform on graph (Kipf and Welling, 2016) and thus where the name graph convolution network comes from.