Graph Representation Learning via Graph Neural Networks
Graph Neural Networks (GNNs) have celebrated many academic and industrial successes in the past years; providing a rich ground for theoretical analysis and achieving state-of-the-art results in several learning tasks. In this talk, I will give an accessible introduction to the area of Graph Representation Learning with a focus on GNNs. I will then give an introductory overview of three recent projects I have worked on. Firstly, we explore the question of how to optimally represent graphs by learning a parametrised graph representation in GNN frameworks. Then, I will discuss work in which we design a GNN to consider neighbourhoods in a graph as a whole, instead of only considering nodes pairwise as is usually done in GNNs. Finally, I will introduce our recent Graph Autoencoder framework which is capable of both detecting communities and predicting links in various real-world graphs, notably including industrial-scale graphs provided by the Deezer music streaming service.