Scalable Graph Neural Network Training

Graph Neural Networks (GNNs) are a new and increasingly popular family of deep neural network architectures to perform learning on graphs. Training them efficiently is challenging due the irregular nature graph data. The problem becomes even more when scaling large graphs that exceed capacity single devices. Standard approaches distributed DNN training, such as data model parallelism, do not directly apply GNNs. Instead, two different have emerged in literature: whole-graph sample-based training. In this paper, we review compare approaches. Scalability with both approaches, but make case research should focus training since it promising approach. Finally, recent systems supporting