This repository introduces DSR students to Geometric Deep Learning on graphs. The material has been compiled, using many different openly available sources, whose list is given below.
- Geometric Deep Learning Lectures: The main source of material, which includes a GDL study book, a course of video lectures, blog posts and videos on the subject.
- Stanford Machine Learning on Graphs Course: The Stanford course, covering many areas of machine learning on graphs.
- Graph Representation Learing: A great introductory book on graph neural networks.
- Graph convolutional networks: The first paper that presented graph convolutional networks.
- Graph attention networks: The paper that demonstrates the graph attention networks.
- Relational inductive biases, deep learning, and graph networks: An amazing DeepMind paper that introduces a general approach for learning on graphs.
- Geometric Dojo Repository: A repo with a practical assignment for the students of a Cambridge Master's program for Representation Learning on Graphs.
- E(n) Equivariant Graph Neural Networks: A paper that introduces E(n) equivariant GNNs, extending the notion of SO(3) equivariance.
- PointNet++: A pioneering paper, introducing the way to use GNNs with point clouds.
- Pytorch Geometric: One of the best libraries for working with graphs.
Other resources can be found inside corresponding Colabs.
- Colab 0 - Introduction to GDL: A brief introduction to the main ideas of GDL.
- Colab 1: Basics of graph theory: Introduction to graphs and their basic properties.
- Colab 2: Graph convolutions: Hands-on guide for graph convolutions.
- Colab 3: PyTorch Geometric and node level tasks: Introduction to PyG via semi-supervised node classification.
- Colab 4: Generalized message passing: Generalized message passing algorithm.
- Colab 0 - Point cloud semantic segmentation: A tutorial on point segmentation, using PointNet++.
- Colab 0 - Equivariant Graph Neural Networks: Practical exercises for understanding invariance and equivariance in GNNs.