[DSC Europe 2020] Applying Graph Neural Networks at the Bleeding Edge - Petar Velickovic

Recent years have seen a surge in research on graph representation learning, including techniques for deep graph embeddings, generalisations of CNNs to graph-structured data, and neural message-passing approaches. I plan to provide a brief overview of the key GNN concepts and architectures, followed by a direct indication of how they've been deployed in challenging real-world contexts. We will cover applications to:
- Computational biochemistry and medicine;
- Traffic simulations and autonomous vehicles;
- Recommendation systems and fake news detection in social networks;
- Understanding, simulating and modelling physical systems.

No prior strong knowledge of GNNs is expected or assumed - only basics of supervised deep learning!

This talk was presented during Data Science Conference Europe 2020 as a part of the Data & AI Research track.