Oxford Certificate Programmes

Generative Modelling - Part of the Artificial Intelligence and Deep Learning course

Two lecture days building generative modelling from its foundations to the current state of the art. The through-line is a single objective — maximum likelihood — and the different ways each model family makes it tractable: energy-based models, variational autoencoders and normalizing flows on day one; diffusion models and flow matching on day two. Slides and the two programming exercises (Google Colab) are below.