Abstract: Graph Convolutional Networks (GCNs) rely heavily on the quality of the input graph, which is often defined by a fixed neighborhood size. This work explores two complementary strategies to ...