Imaging in Paris Seminar

Parisian Seminar on the Mathematics of Imaging

Welcome to the website of the Parisian Seminar on the Mathematics of Imaging !

The goal of this seminar is to cover the fields of the mathematics of imaging in a very wide sense (including for instance signal processing, image processing, computer graphics, computer vision, various applications and connections with statistics and machine learning). It is open to everyone. It takes place at Institut Henri Poincaré on the **second (third in Sept./Nov./Apr.) Tuesday of each month from 2pm to 4pm. Each seminar is composed of two presentations.

You can subscribe or unsubscribe to the mailing list of the seminar and to the agenda of the seminar.

Upcoming seminars

Click on the title to read the abstract.

Lisl Weynans (Institut de Mathématiques de Bordeaux)
October 13th, 2pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

Maxime Di Folco (Télécom Paris)
November 17th, 2pm, Amphi Choquet-Bruhat (Perrin building).
Title: Multimodal Representation Learning for Clinical Decision Support from Vision and Tabular Data ⬇
 Abstract: Clinical decision support utilising multimodal AI has advanced significantly in recent years, transforming medical imaging from a standalone diagnostic visual tool into an active, predictive anchor within modern healthcare. However, seamlessly fusing high-dimensional imaging modalities with complementary clinical data remains a central challenge due to severe data heterogeneity and representation mismatches. In this talk, I will present recent imaging-centric vision-tabular approaches applied to cardiovascular diseases, and show how addressing these specific fusion hurdles enables us to challenge established clinical baselines and drive better decision-making for patients.

TBA
November 17th, 3pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
December 8th, 2pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
December 8th, 3pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
January 12th, 2pm, Room Yvette Cauchoix (Perrin building).
Title: ⬇
 Abstract:

TBA
January 12th, 3pm, Room Yvette Cauchoix (Perrin building).
Title: ⬇
 Abstract:

TBA
February 9th, 2pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
February 9th, 3pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
March 9th, 2pm, Room Yvette Cauchoix (Perrin building).
Title: ⬇
 Abstract:

TBA
March 9th, 3pm, Room Yvette Cauchoix (Perrin building).
Title: ⬇
 Abstract:

TBA
April 20th, 2pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
April 20th, 3pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
May 11th, 2pm, Room Yvette Cauchoix (Perrin building).
Title: ⬇
 Abstract:

TBA
May 11th, 3pm, Room Yvette Cauchoix (Perrin building).
Title: ⬇
 Abstract:

TBA
June 8th, 2pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

TBA
June 8th, 3pm, Amphi Choquet-Bruhat (Perrin building).
Title: ⬇
 Abstract:

Previous seminars of 2026-2027

The list of seminars prior to summer 2026 is available here.

Guy Gilboa (Technion)
September 15th, 2pm, room SCAI seminar room - Esclangon building, 1st floor, Campus Pierre et Marie Curie, 4 place Jussieu, Paris.
Title: How to Encode World Knowledge? ⬇
Abstract: Foundation models are a key platform which implicitly encodes world knowledge. In this talk we first focus on vision-language models, such as CLIP, and investigate their geometric behavior and logic behind the high-dimensional feature encoding. For instance, we find that as image or text become more rare and distinct they are encoded further from the center of the embedding. We explain why InfoNCE loss leads to that behavior. We also find out empirically that each modality can be well modeled statistically as admitting a multivariate Gaussian distribution. This finding is then proved formally, where we show InfoNCE asymptotically induces Gaussian statistics. Finally, we connect two major image foundation models – encoders and generators, through a universal normal embedding hypothesis. A surprising consequence of this hypothesis is demonstrated, where generative diffusion “noise” contains semantic data which can be accessed by linear probing for either classification or editing. This talk is based on 6 recent papers in the group published at ICML 2025, ICLR 2026 and CVPR 2026.

Organizers

- [Jonathan Vacher](https://jonathanvacher.github.io/) (MAP5 @ Université Paris Cité)

Thanks

The seminar is hosted by IHP, and supported by RT-MAIAGES, Télécom Paris and CMM Mines Paris - PSL.