Geometric Deep Learning course, Lecture Notes

To access the document, please click [PDF]

190 pages of lecture notes for the fast evolving field of Geometric Deep Learning. Written to explore the topic and shape research interest under the suggestion of mentors from the Visiting Student initiative. The course, the slides and the book are available here

Abstract

Geometric Deep Learning (GDL) is a fast emerging field of research. Where Deep Learning fails to present a unifying framework, GDL is a valid proposal for a theoretical framework under high dimensional learning. This document is a collection (hopefully in continuous expansion) of the Lecture Notes of the course held at the 2021 African’s Master in Machine Intelligence (AMMI). Recordings are available online, and a book is currently being written. To craft such a framework notions from Statistics, Group Theory, Geometry, Fourier Analysis, Continuous and Discrete Spaces are beautifully aligned together. The result is a principled and contained collection of basics from which many architectures can be derived. Up to my knowledge, this is the first set of lecture notes from the course. In the future, I plan to complete it in its entirety, and enlarge it with deeper theoretical insights on the methods referenced, or with the help of additional material available, and an upcoming Summer School in Pescara, Italy.
Disclaimer 1 The document could and might be expanded in the future.
Disclaimer 2 Chapters are in the order of teaching. Images are entirely taken from the presentations, and come from different sources, please refer to the slides to find the exact origins.