This course will consist of presentations, live coding, and Q&A sessions.

The video call invitation (Google Meet) will be provided on the 27th of May.

Graphs are universal data structures that can represent complex relational data. This course will explore and try to explain the most important modern graph neural networks and computational modules. We will learn how to build a GNN from scratch, walking through each part of the model and explain it. In the second part of the course, we will bring convolution to graph models, and we will explore Graph Convolutional Networks (GCN), Graph Attention Networks (GAT), Graph Sample and Aggregate (GraphSAGE) and Graph Isomorphism Network (GIN). The course will also cover the challenges of using graphs in machine learning, how to overcome them and some empirical GNN design lessons.