ATML: Advanced Topics in Machine Learning

Welcome

Welcome to the course webpage for Advanced Topics in Machine Learning. This course is designed for students who aspire to become technical experts, pursue research, and innovate in the field of machine learning.

This page will be updated with further information about the course, including the syllabus, schedule, and other resources.

Learning outcomes

  • Identify how an aspect of an advanced machine learning topic applies to a given applied problem.
  • Derive mathematical details of machine learning methods in the topic area.
  • Critically compare and contrast alternative choices or variants of methods or approaches in the area.
  • Create accessible and useful explanations of the workings and failure modes of machine learning methods, including appropriate mathematical and implementation detail.
  • Identify the ethical and societal implications, including both benefits and risks, of the deployment of machine learning methods in the area.

Lecture recordings

All lecture recordings should be accessed via Learn; you will need to log in using your EASE account. (Learn provides you with access to any lecture recordings available for this course. You will need to select the "lecture recording" link once, before you can access any direct links to a lecture recording.)

Topic tracks

Three tracks will be offered in the course in 2026. It is expected that students will follow two out of the three tracks.

Geometric learning (track page)

Topics: graph neural networks, models on non-Euclidean domains (e.g., manifolds such as spheres), symmetries, invariances and equivariances.

Instructor: Viacheslav Borovitskiy

Deep generative modelling (track page)

Topics: learning and inference in families of neural probabilistic models used for high-dimensional data: deep latent variable models (e.g., autoencoders), generative adversarial models, normalising flows, diffusion models. Principles of representation learning, generalisation, model assessment and deployment considerations.

Instructor: Nikolay Malkin

Optimisation in machine learning (track page)

Topics: non-convex optimisation and neural networks, loss landscapes in high dimensions, algorithmic stability, overfitting and overparameterisation in neural networks.

Instructor: Rik Sarkar

License
All rights reserved The University of Edinburgh