Introduction

This training workshop provides an introduction to the mathematical foundations of deep learning and its practical applications, with a specific focus on solving ill-posed inverse problems.

Topics covered include:

• Foundations of deep learning: neural network architectures, activation functions, training strategies, backpropagation
• Bilevel optimisation and algorithm unrolling for learned regularisation
• Plug-and-Play priors and convergence theory
• Generative models, diffusion models, and deep equilibrium networks

The workshop consists of morning lectures, afternoon coding sessions, and group presentations on Friday. Students will work in small groups on an image reconstruction challenge, competing to achieve the best PSNR and SSIM on a hidden test set.

Lecturers

Prof. Martin Benning, University College London, Department of Computer Science

Dr Riccardo Barbano, University College London, Department of Computer Science