Rencontre, 10 December 2026
Nous serons en salle 435 le matin, et en amphi A l'après-midi. La salle 435 est au 4ème étage, sur le site Monod, dans l'aile opposée par rapport à l'amphi B. L'amphi A est au 3ème étage, à l'autre bout du couloir par rapport à l'amphi B.
La rencontre sera diffusée en ligne, des instructions seront disponibles sur cette page peu de temps avant la rencontre.
Programme
-
Arthur Gall (National Institute of Informatics and SOKENDAI (The Graduate University for Advanced Studies) (JP))
In this talk, I will present how the modeling of parametric Markov chains via categories with finite biproducts led to interesting theoretical results, such as a compositional notion of Jacobian, as well as concrete algorithmic payoffs. In particular, this perspective yields a fast and sound differentiation algorithm that outperforms verification solvers on certain tasks. This talk is based on a recent preprint (https://arxiv.org/abs/2610.36655) and extends a previous talk (https://sites.google.com/view/differential-kyoto-2026) with more theoretical results.
-
Ivan Di Liberti (Göteborgs universitet (SE))Invited talk: Craig interpolation for subgeometric logics
We show that a vast class of finitary fragments of geometric logic admit a form of Craig interpolation property. In doing so, we provide a new dictionary to import technology from algebraic logic to categorical logic. The talk is based on https://arxiv.org/abs/2601.11221 and https://arxiv.org/abs/2610.29771.
-
Pawel Sobocinski (Tallinn University of Technology (ES))Invited talk: Diagrammatic First-Order Logic
First-order bicategories are categorical structures that capture classical first-order logic with equality, including models with empty domains. They come with a string-diagrammatic syntax closely related to the Beta system of Peirce’s existential graphs, and with a sound and complete axiomatisation. I will introduce first-order bicategories and show how they support a Lawvere-style functorial semantics: models of a theory are morphisms out of the free first-order bicategory it generates, and the natural 2-cells between models turn out to be exactly the elementary embeddings. Morphisms between theories are subtler. Classical interpretations, which build models of one theory from definable sets and definable quotients of another, correspond to morphisms into a free completion obtained by splitting partial equivalence relations. In this setting, two theories are bi-interpretable, the model theorist’s standard notion of sameness, if and only if their completions are equivalent.
Joint work with Filippo Bonchi, Alessandro Di Giorgio, Roberto Di Virgilio and Nathan Haydon.
- 10 December 2026, 10:30



