70105

Deep Graph-Based Learning

Module aims

This module comes at the intersection of graph theory, machine learning, and deep modeling. It aims to give a flavor of different aspects of the broad and ever-expanding field of graph theory and learning. The module first motivates the need for graph-based learning, by introducing conventional graph data analysis methods. Next, the module covers the nascent field of graph neural networks on both theoretical and application levels. Together, we will explore the foundation of landmark GNN models and cover state-of-the-art models and graph-based learning techniques. We will have sessions to implement a few GNN examples in Python, weekly quizzes for continuous assessment, and stimulating group discussions based on deconstructing recent papers published in this field.

Learning outcomes

Upon successful completion of this module you will be able to:
1. Evaluate the relative merits of graph learning and traditional graph data analysis.
2. Analyze the design choices for a Graph Neural Network (GNN) architecture.
3. Implement GNN models using state-of-the art programming languages and tools.
4. Apply graph-based learning to real-world problems in image/text/speech processing and graph analysis.

Module syllabus

1. Graph fundamentals and topological principles
2. Traditional graph-based analysis methods
3. Graph convolutional networks (GCN)
4. Theory of graph neural networks (GNNs)
5. Applications of GNNs
6. Generative and diffusion models for GNNs
7. Data-specific challenges in GNNs
8. The future of deep graph learning

Teaching methods

The material will be taught through lectures backed up by exciting formative group studies and coding tasks, each with a design thinking and algorithmic perspective. You will receive technical support for the various coding tasks from Graduate Teaching Assistants (GTAs).

An online service will be used as a discussion forum for the module.

Assessments

There will be one coursework exercise contributing 30% and a project contributing 40% of the marks. There will be weekly timed quizzes contributing 30% of the marks, which will test both theoretical and practical aspects of the subject.

You will get live feedback on your solutions to the formative in-class problems during the session.

The answers to the weekly quiz will be provided following each quiz, with explanations. You will receive written feedback on each coursework submission, following a marking scheme set by the module leader. This feedback explains where marks were lost, and how they could have been obtained. General feedback on the coursework will also be provided during the lectures.

There will be detailed written feedback for each of the assessed exercises and class-wide feedback explaining common pitfalls and suggestions for improvement. For unassessed tutorials with the lecturers and GTAs, written solutions will be provided after an in-class tutorial session.

Module leaders