We go through steps in creating GNNs with DGL. We will cover common graph convolution layers including GCN, SageConv, GAT. In each case, we implement the convolution layer from scratch and compare it with the built-in function from DGL.
- GNN with GCN
- GNN with GraphSage
- Link prediction with GraphSage
- Graph Classification
- Graph Attention Network (GAT)
Most of the materials are taken from DGL documentation https://docs.dgl.ai/tutorials/blitz/index.html