In representation learning on graphs, graph neural networks (GNNs) have been widely employed and have attained
cutting-edge performance in tasks like node categorization and link prediction. However, the majority of GNNs now in use are made to learn node representations on homogenous and fixed graphs.