This repository provides implementations of two graph convolutional architectures: MixHop (ICML 2019) and N-GCN (NeurIPS 2018). MixHop layers mix multiple powers of the adjacency matrix to learn delta operators, while N-GCN provides a higher-order graph convolutional layer. Both implementations maintain the same memory footprint and computational complexity as standard GCN while capturing multi-scale neighborhood information for improved node classification on graph data.