Café π news

To celebrate British Sandwich Week (19–25 May), we’re introducing a special baguette in the Café: 

The Coronation Chicken Baguette.

Erlangen AI Hub Conference 2025

The Erlangen AI Hub Conference will bring together leading minds from across the UK’s mathematical, algorithmic and computational communities to advance the application of pure mathematics in AI. 

9-11 June 2025

Maths Building, Queen Mary University of London, Mile End Road, London E1 4NS  

Registration deadline: 23 May 2025 (12 noon). Link above. 

Image: Leonardo da Vinci - Study of Arms

Topological model selection: a case-study in tumour-induced angiogenesis
 McDonald, R Byrne, H Harrington, H Thorne, T Stolz, B (21 Apr 2025)

3 Minute Thesis competition

Present your research in just three minutes to win a prize. Oxford’s local SIAM-IMA student chapter invites you to give a three minute talk with the aid of a single slide aimed at a non-specialist audience. This competition is open to ALL research students in the Mathematical Institute, in both pure and applied mathematics.

Drop rebound at low Weber number
 Gabbard, C Aguero, E Cimpeanu, R Kuehr, K Silver, E Barotta, J Galeano-Rios, C Harris, D (01 May 2025) http://arxiv.org/abs/2505.00902v1
Fri, 20 Jun 2025

12:00 - 13:00
Quillen Room

TBD

Mario Marcos Losada
(University of Oxford)
Abstract

TBD

On the rectifiability of $\mathsf{CD}(K,N)$ and $\mathsf{MCP}(K,N)$
spaces with unique tangents
 Magnabosco, M Mondino, A Rossi, T (02 May 2025) http://arxiv.org/abs/2505.01151v1

Talking Maths in Public

This year, the Talking Maths in Public conference will take place at the University of Warwick and online, on Thursday 28th - Saturday 30th August. TMiP is a biannual meeting for people who communicate maths in a variety of forms, from professional outreach providers to people who deliver maths enrichment activities alongside their work. 

Fri, 16 May 2025
13:00
L6

Certifying robustness via topological representations

Andrea Guidolin
(University of Southampton)

The join button will be published 30 minutes before the seminar starts (login required).

Abstract
Deep learning models are known to be vulnerable to small malicious perturbations producing so-called adversarial examples. Vulnerability to adversarial examples is of particular concern in the case of models developed to operate in security- and safety-critical situations. As a consequence, the study of robustness properties of deep learning models has recently attracted significant attention.

In this talk we discuss how the stability results for the invariants of Topological Data Analysis can be exploited to design machine learning models with robustness guarantees. We propose a neural network architecture that can learn discriminative geometric representations of data from persistence diagrams. The learned representations enjoy Lipschitz stability with a controllable Lipschitz constant. In adversarial learning, this stability can be used to certify robustness for samples in a dataset, as we demonstrate on synthetic data.
Subscribe to