Introduction to Geometric Deep Learning

The use of deep learning has grown exponentially in the last decade, and a diverse set of architectures has been developed in computer vision, natural language processing, etc. Geometric deep learning is a method to understand the underlying and unifying symmetries and geometry between the various neural network architectures. In this talk, we will discuss the basic ideas of deep learning and how several core architectures such as CNNs, GNNs, and transformers naturally arise from a geometric perspective.