Mathematical Foundations
of Deep Learning

3 – 7 August, 2026

A one-week summer school providing rigorous education in the foundational mathematics of modern AI at the Centre for Mathematical Sciences (Wilberforce Road, Cambridge CB3 0WA, UK).

Rigour in the Age of AI

This summer school features four day-long lecture courses addressing topics often fragmented across different university departments. We are hosting graduate students and early career researchers to favour tight-knit interactions and collaborative dialogue.

Supported by Maths4DL and ProbAI.

Scientific Programme

1

Approximation Theory

Key results on neural network approximation, generalised to neural operators for learning mappings between infinite-dimensional function spaces.

2

Generative Modelling

Exploration of flow matching as a bridge between ODEs and density estimation, offering a deterministic alternative to diffusion models.

3

Structure Preserving Deep Learning

Designing deep learning methods that preserve geometric and dynamical structure, with foundations in numerical analysis, differential equations, and structure-preserving discretisation.

4

Inverse Problems & Optimisation

Addressing complex optimisation challenges, drawing connections between classical methods and modern deep learning frameworks.

Structure of the Week

Each lecturer will teach for one day, roughly from 09:00 to 17:00.
Wednesday morning is dedicated to participant talks/posters, followed by a free afternoon to explore Cambridge.

Tentative Schedule

Time Monday Tuesday Wednesday Thursday Friday
Morning Eldad Haber Nikola B. Kovachki Short Talks & Posters Audrey Repetti Brynjulf Owren
Afternoon Eldad Haber Nikola B. Kovachki Free Afternoon Audrey Repetti Brynjulf Owren

Lecturers

Nikola B. Kovachki

NVIDIA / Caltech

Nikola B. Kovachki


Expert in approximation theory for infinite-dimensional spaces.

Eldad Haber

UBC

Eldad Haber


Leader in mathematical modelling of deep learning frameworks.

Brynjulf Owren

NTNU

Brynjulf Owren


Researching structure-preserving methods for deep learning and differential equations.

Audrey Repetti

Heriot-Watt

Audrey Repetti


Leading researcher in inverse problems and continuous optimisation.

Organising Committee

Prof. Carola-Bibiane Schönlieb

University of Cambridge

Prof. Chris Budd OBE

University of Bath

Dr Moshe Eliasof

University of Cambridge

Dr Davide Murari

University of Cambridge

Cambridge

Logistics & Fees

To foster interaction, we encourage attendees to present their own research on Wednesday morning.

Registration Fee: £100

Includes lunch & coffee breaks on Mon, Tue, Thu, Fri, plus Wed morning coffee. (Dinners are not included).

Link to registration payment coming soon

Location & Transport

The Centre for Mathematical Sciences is located on Wilberforce Road (CB3 0WA). It is west of the city centre and easily accessible by a 15-minute walk, cycling, or the regular U-Bus service from the centre.

Accommodation

Students must book their own accommodation promptly. We recommend booking college rooms via University Rooms. Churchill College, Fitzwilliam College, and Robinson College are the closest to the CMS.

Free Afternoon

Wednesday afternoon is kept completely free for participants to explore the historic city of Cambridge.

Admissions 2026

Applications Closed

The application deadline has now passed.

Results on the selection will be shared by the end of May / beginning of June.

Thank you to everyone who applied.

The application form is no longer accepting submissions.