Doctorants

A first simplification of the gene expression mechanism considers that a gene is transcribed into messenger RNA, which in turn is translated into protein. Single-cell data have revealed the presence of biological variability between cells of identical genome and environment, highlighting not only epigenetic aspects but also the stochastic nature of gene expression.

In the context of regulatory networks underlying cell states and types, we need to build a model that takes into account both stochasticity and the interaction of genes with each other. Here we focus on a dynamical model of gene expression, formulated as a piecewise-deterministic Markov process (PDMP) and describing an arbitrary number of interacting genes. This stochastic model is able to reproduce the biological variability measured experimentally, but remains mathematically complex and difficult to study. This is why, in the litterature, a simplified model with only proteins is considered. 

During this talk, we provide insights on construction and use of semigroups and infinitesimal generators for PDMPs. Afterwards we present both models and use coupling methods to explicitly upper bound the error made when substituting the full model with its simplified version.

TBA

TBA

TBA

TBA

TBA

L’objectif de cette journée est de rassembler les doctorants de Nancy et de Metz afin de faire plus ample connaissance autour d’exposés mathématiques.

Le programme est de 3 exposés le matin et 3 exposés le soir. La journée se passera entièrement à l’Amphi 7.

Organisateurs: Nicolas Frantz (Metz), Jimmy Payet (Metz) et Pierre Popoli (Nancy).

Les variétés de Siegel sont des variétés de Shimura qui paramètrent des variétés abéliennes avec une polarisation. Le premier exemple est la courbe modulaire dont l’importance est cruciale en théorie des nombres : elle intervient dans la preuve du théorème de Fermat-Wiles et plus généralement dans la correspondance de Langlands pour $G{L}_2$ sur $\mathbb{Q}$. Dans cet exposé, j’introduirai les variétés de Siegel en tant que variétés algébriques sur un corps fini et je décrirai les propriétés géométriques de certains fibrés vectoriels automorphes vivant dessus.

Les opérateurs de Schrödinger sont des incontournables dans la mécanique
quantique. J’exposerai d’abord des motivations physiques de l’étude
spectrale de ces objets. Plusieurs auteurs ont obtenu des bornes en
norme $L^p$ sur les quasi-modes des opérateurs de Schrödinger. On verra
ensuite comment se généralisent de telles estimées à des systèmes
orthonormés de fonctions. L’idée de l’exposé est de donner un avant-goût
des jolis outils sous-jacents.

Résumé à venir

In modern algebraic geometry, the study of moduli spaces plays a central role in the problem of classifying certain geometric objects (e.g., Riemann surfaces, vector bundles), up to a fixed notion of isomorphism. The foremost question arising is whether we can construct a moduli space which, roughly speaking, parametrizes the isomorphism classes of such objects. The moduli space will be endowed with a natural geometric structure, which is often a scheme or an algebraic stack. In this talk we give an introduction in the theory of moduli spaces, with special emphasis on some classical examples: the Grassmannian, the Hilbert scheme, the moduli space of sheaves etc.. We will formulate the moduli problems using the categorical language of representable functors, and introduce the notions of fine and coarse moduli spaces.

Stochastic Approximation is a useful tool for Machine Learning techniques such as Stochastic Gradient Descent. These algorithms are applied to a lot of different fields, improving the transportation times, helping doctors diagnosing with medical images, automatically translating text, detecting spam and more. Most of the time, the model traditionally lies in a vector space. However, some problems present non-linear constraints, that can be translated into a manifold. This framework ensures the conservation of key properties. As an introduction to geometric machine learning, we study the gradient descent algorithm, and its adaptation to Riemannian manifolds. Finally, we compare the performance of the two, introducing new non-asymptotic bounds.