Maria Bånkestad: Geometric deep learning with graph neural networks

Geometric deep learning with graph neural networks is the title of this week's Machine Learning Seminar by Maria Bånkestad. Deep learning has delivered outstanding results in applications with an underlying Euclidean structure. Maria describes the principles of graph neural networks which has recently emerged as the, perhaps, most promising method for non-euclidean appliactions like biology, physics, and computer graphics.