README.md

Deep Graph Learning (DGL, 2024)

Taught by Prof. Islem Rekik at Imperial College London

---

Introduction

This repo contains all the lecture notes for this DGL course. All relevant records for this course can be accessed at BASIRA Lab.

---

Course Lectures

• Lecture 1:

  • Lecture 1.1: Graph types

  • Lecture 1.2: The Graph matrix

  • Lecture 1.3: Graph learning tasks

• Lecture 2:

  • Lecture 2.1: The logic behind graph-based learning

  • Lecture 2.2: The evolving landscope of feature embedding

  • Lecture 2.3: Shallow graph node embedding

  • Lecture 2.4: Analyzing a single GCN layer

  • Lecture 2.5: Generalized GCN node and layer updates

• Lecture 3:

  • Lecture 3.1: GCN training and loss optimization

  • Lecture 3.2: GNN inductive capability & graph-based learning

  • Lecture 3.3: Graph pooling & embedding aggregating

  • Lecture 3.4: GCN layer operations

  • Lecture 3.5: Global and local aggregation methods

• Lecture 4:

  • Lecture 4.1: Point, batch and mini-batch gradient descent

  • Lecture 4.2: Batching and GNN sampling methods

  • Lecture 4.3: Recap on GNN sampling methods

  • Lecture 4.4: GNN batch normalization layer

  • Lecture 4.5: Generalized GNN layer and Dropout

  • Lecture 4.6: GNN inductive vs transductive learning

• Lecture 5:

  • Lecture 5.1: Node permutation invariance in GNNs

  • Lecture 5.2: Node permutation equivariance in GNNs

  • Lecture 5.3: GNN expressiveness

  • Lecture 5.4: Graph Isomorphism Network Expressive Nets

• Lecture 6:

  • Lecture 6.1: Overview of supervised generative GNNs

  • Lecture 6.2: Self-supervised/unsupervised generative GNNs

  • Lecture 6.3: Unconditional sequential graph generation

  • Lecture 6.4: Unconditional one-shot graph generation

  • Lecture 6.5: Supervised conditional generation on graphs

  • Lecture 6.6: Generative Graph U-Net

  • Lecture 6.7: Evaluation measures for generative GNNs

---

Homeworks

---

Paper analysis worksheets

---

Project

---

Tutorials

---