Table des matières


Séminaire

Manage talks

Stochastics and Biology seminar


Day, hour and place

1st Wednesday of each month at 11:00,


Contact(s)

Catherine Matias


Next talks

Stochastics and Biology seminar
Wednesday June 10, 2026, 11AM, 16-26.209
Aurélien Velleret On the LAN and LAMN Properties for Mean-Field Model of Interacting Neurons

We study a mean-field system of interacting particles represented by stochastic differential equations (SDE's) with jumps, introduced in [1] as a model describing the time evolution of neuronal potentials in the human brain. Depending on the renormalization of the interaction : either averaging scaling (1/N) or diffusive scaling (1/\sqrt N)-it was shown in [1] and [2] that, as the number of particles tends to infinity, the system exhibit the propagation of chaos and conditional propagation of chaos, respectively. Assuming that the jump intensity depends on an unknown parameter, we establish the Local Asymptotic Normality (LAN) property in the averaging case and the Local Asymptotic Mixed Normality (LAMN) in the diffusive case. Some implications regarding the performance of the maximum likelihood estimator are also derived. The limiting Fisher information depends on the limit of the empirical measure of the system. It is deterministic in the averaging regime, random in diffusive regime. This phenomenon explains the dichotomy between LAN and LAMN. Based on joint works with Xavier Erny, Aline Duarte, Eva Löcherbach, Dasha Loukianova.

[1] Fournier, N. and Löcherbach, E. (2016). On a toy model of interacting neurons. Ann.Inst. Henri Poincaré Probab. Stat. 52(4), 1844–1876. [2] Erny, X., Löcherbach, E. and Loukianova, D. (2021). Conditional propagation of chaos for mean field systems of interacting neurons. Electron. J. Probab. 26, 1–25.

Stochastics and Biology seminar
Wednesday July 1, 2026, 11AM, 16-26.209
Barbara Bricout (LPSM (MAV)) To be announced.

To add the talks calendar to your agenda, subscribe to this calendar by using this link.


Previous talks


Year 2026

Stochastics and Biology seminar
Wednesday May 6, 2026, 11AM, 16-26.209
Alexandre Chaussard (LPSM (MAV)) Structured Latent Generative Models and Variational Inference for Microbiome Analysis

The human gut microbiome is a complex ecosystem whose composition is derived from high-throughput sequencing technologies, yielding hierarchical count data indexed by microbial taxa. These microbiome profiles offer major opportunities for understanding host-microbes interactions, yet they come with substantial statistical challenges: counts are discrete, sparse, and overdispersed; sequencing induces compositional constraints and high-dimensional profiles, while clinical applications often take place in heterogeneous and small cohorts, with imperfect labeling regarding the underlying biology.

In this thesis, we develop structured latent generative models for microbiome count data and apply them to clinical studies. Our methodological contributions notably rely on Poisson log-normal (PLN) models which provide a principled probabilistic framework tailored to multivariate counts, and on variational inference which enables scalable learning while leveraging data structure. First, we incorporate the hierarchical organization of microbial taxa by introducing a tree-based extension of PLN models and by establishing identifiability results that support principled interpretation. We then build on this framework to propose a model-based data augmentation strategy that enhances predictive performance across clinical tasks while preserving ecological coherence. Subsequently, we extend our latent generative viewpoint to longitudinal settings through a perturbation-aware independent component analysis model for temporal count data, joined with identifiability guarantees allowing principled interpretation of the inferred components and regimes. Finally, we consider two inflammatory bowel disease case studies to highlight the clinical constraints that shape microbiome analyses, and to illustrate how parts of our methodology can be leveraged for prognosis in a statistically challenging setting.

Overall, this thesis argues that latent generative modeling offers a compelling framework for microbiome analysis, enabling the incorporation of biological structure while connecting representation learning, preprocessing, and data augmentation within a single probabilistic perspective. In particular, interpretability in latent microbiome models hinges on identifiability, and introducing structural information turns probabilistic models into principled tools for extracting meaningful representations and enhancing the statistical power of microbiome profiles.

Stochastics and Biology seminar
Wednesday April 1, 2026, 11AM, 16-26.209
Annulé annulé

Stochastics and Biology seminar
Wednesday March 4, 2026, 11AM, 16-26.209
Maxime Egéa (LPSM) Local differential privacy in survival analysis using private failure indicators

Survival analysis is often applied to sensitive data making privacy protection of crucial importance. In this talk, we consider local differential privacy in survival analysis using private failure indicators. First, I will provide an overview of existing methods and discuss how individual privacy can be preserved. Next, I will introduce a non-parametric kernel estimator for the cumulative hazard function using a locally differentially private mechanism on failure indicators. Under mild conditions, we also prove lower bounds on the minimax rates of convergence and show that our estimator is minimax optimal under well-chosen bandwidths. Finally, I will illustrate our estimator with numerical results.

Stochastics and Biology seminar
Wednesday February 4, 2026, 11AM, 16-26.209
Alexander Reisach (Univ Paris saclay) The Promise and Pitfalls of Causal Graphs

Stochastics and Biology seminar
Wednesday January 7, 2026, 11AM, 16-26.209
Valentin Schmutz (Univ. College London) Concentration of measure in “low-rank” biological neural networks

Recurrent neural networks with low-rank connectivity matrices are general and tractable models of collective dynamics in large networks. This class of models dates back to the seminal works of J. Hopfield (1982) and S. Amari (1972), and it still plays an instrumental role in computational neuroscience today. To highlight the analytical tractability of these models, I will first review some recent theoretical results concerning the case where the low-rank connectivity is random, the rank is kept fixed, and the number of neurons tends to infinity. In this case, we find that (i) the dynamics of the network converges to a neural field equation, (ii) the dynamics can be reduced to a latent, low-dimensional dynamical system, and (iii) the latent dynamics can be fully solved in certain special cases. In the second part of the presentation, I will show that low-rank connectivity is associated with remarkable concentration of measure phenomena in networks of biological neurons. Considering a feedforward network setup where neurons in the first layer, modelled as Cox processes, transmit stochastic spikes, I will present a theorem stating that the network can behave as if each spiking neuron were transmitting its subthreshold membrane potential as both the rank of the connectivity and the number of neurons tend to infinity. This result could explain how, at the network level, neurons can transmit their subthreshold membrane potential fluctuations through sparse spikes. The proof of the theorem involves the so-called thin shell phenomenon, a well-known concentration phenomenon in high-dimensional probability.


Year 2025

Stochastics and Biology seminar
Wednesday December 3, 2025, 11AM, 16-26.209
Léo Micollet (LPSM (MAV)) Stochastic Models for Mosquito Population Dynamics and Control Using the Sterile Insect Technique

Stochastics and Biology seminar
Wednesday November 5, 2025, 11AM, 16-26.209
Adrien Cotil (LJLL) Une approche probabiliste pour l'étude du flocking du modèle de Cucker-Smale

Le modèle de Cucker-Smale est un modèle largement utilisé pour décrire le comportement de groupes d'animaux tels que les nuées d'étourneaux, les bancs de poissons ou les essaims d'insectes. Il décrit des individus qui tendent à se déplacer dans la même direction et à la même vitesse que leurs voisins, correspondant biologiquement à un principe de mimétisme. L'une des principales questions mathématiques posées par ce modèle est la détermination conditions suffisantes sur les paramètres et les données initiales garantissant l'apparition d'un consensus global, appelé “flocking” dans ce contexte. Dans le cas d'individus grégaires, comme les bovins ou les ovins, les interactions entre individus peuvent être asymétriques, en raison de la spécialisation de certains individus dans certaines fonctions comme la conduite du troupeau ou l'exploration du territoire. La prise en compte de cette asymétrie dans le modèle de Cucker-Smale entraîne la perte de l'échangeabilité des individus, qui est fortement utilisée pour établir des conditions de flocking dans le cadre classique. Dans cet exposé, nous aborderons une nouvelle approche pour l'étude du flocking du modèle de Cucker-Smale non échangeable, basée sur une interprétation probabiliste de ses solutions. Cette approche permet de ramener la question du flocking à la convergence en loi d'un processus de saut markovien inhomogène en temps. L'étude de cette convergence se fera au moyen du coefficient d'ergodicité de Dobrushin, qui permet de quantifier le taux de contraction d'un semi-groupe markovien vers son sous-espace vectoriel invariant.

Stochastics and Biology seminar
Wednesday October 1, 2025, 11AM, 16-26.209
Bixuan Liu (LPSM) Identifiability of VAR(1) model in a stationary setting

Stochastics and Biology seminar
Wednesday July 2, 2025, 11:30AM, 15-16.309
Laura Kanzler (CNRS - LJLL) Modelling the evolution of the size-distribution in aquatic ecosystems

Trophic interactions between animals in the ocean were matter of interest since the 1960ies, where it was quickly discovered that the body size of individuals acts as ’master trait’ in food webs of animals, giving rise to emergent distributions of biomass, abundance and production of organisms. We propose and investigate a deterministic jump-growth model, which is given by a kinetic equation for coalescing particles, aiming to capture this emergence phenomenon in aquatic ecosystems. The equation of interest is derived from individual based dynamics governed by a stochastic process. Following the observation of the body mass being the crucial trait in these dynamics it is based on the assumption that binary interactions between individuals in the ecosystem take place: A predator feeding on a prey, which then results in growth of the predator with assimilating a certain (usually very small) amount of its prey’s mass as well as plankton production. Analytical results in various parameter regimes are discussed and numerical simulations underlying these observations are given.

Stochastics and Biology seminar
Wednesday June 11, 2025, 11AM, 16-26.209
Roland Sogan (LPSM) Sparse Gaussian Graphical Models with Latent Cluster Structure

Stochastics and Biology seminar
Wednesday May 7, 2025, 11AM, 16-26.209
Sarah Kaakai (Université Sorbonne Paris Nord) Non-parametric estimation of the smurf transition rate and mortality rate in a two-phase model of aging

Stochastics and Biology seminar
Wednesday April 2, 2025, 11AM, 16-26.209
Charlotte Dion-Blanc (LPSM) Processus de Hawkes

Stochastics and Biology seminar
Wednesday March 5, 2025, 11AM, 16-26.209
Diarra Fall (Université d'Orléans (en délégation au LPSM)) Nonparametric Bayesian methods for reconstructing spatial (3D) and space-time (3D+t) Positron Emission Tomography (PET) images.

Stochastics and Biology seminar
Wednesday February 5, 2025, 11AM, 16-26.209
Madeleine Kubasch (Sorbonne Université - Chaire MMB) Large Population Limit for a Multilayer SIR Model Including Households and Workplaces

Stochastics and Biology seminar
Wednesday January 8, 2025, 11AM, 16-26.209
Elhacène Djaout (LPSM) Modélisation épidémiologique pour les eaux usées - Obépine


Year 2024

Stochastics and Biology seminar
Wednesday December 4, 2024, 11AM, 16-26.209
Arnaud Liehrmann (Sorbonne Université - LCQB) DiffSegR:An RNA-Seq data driven method for differential expression analysis using changepoint detection

Stochastics and Biology seminar
Wednesday November 6, 2024, 11AM, 16-26.209
Sarah Ouadah (LPSM) Présentation thèmes de recherche

Stochastics and Biology seminar
Wednesday July 3, 2024, 11AM, 16-26.209
Stéphane Robin (LPSM) Réseaux plantes-pollinisateurs et motifs

Stochastics and Biology seminar
Wednesday June 5, 2024, 11AM, 16-26.209
Michèle Thieullen (LPSM) Étude de modèles à conductances

Stochastics and Biology seminar
Wednesday May 15, 2024, 11AM, 16-26.209
Grégory Nuel (LPSM) Méthode INLA (integrated nested Laplace approximation) en approximation bayésienne

Stochastics and Biology seminar
Wednesday April 3, 2024, 11AM, 16-26.209
Rémi Boutin (LPSM) The Deep Latent Position Topic Model

Stochastics and Biology seminar
Wednesday March 6, 2024, 11AM, 16-26.209
Catherine Matias (LPSM) Clustering dans les hypergraphes


Year 2022

Stochastics and Biology seminar
Tuesday November 29, 2022, 11AM, Salle Jacques Neveu, 16-26-113
Aurélien Velleret (LAMA, Marne-la-vallée) Infections inter-villes et prise en compte de leurs tailles très hétérogènes dans les stratégies de confinement

Au cours des différentes vagues de propagation du covid dans des pays comme la France ou la Pologne, nous avons pu constater que les grandes villes étaient les plus rapidement touchées, ce qui suggère que le niveau élevé de connexions entre elles a joué un rôle accélérateur décisif. Suivant ce qui a été reconnu comme la loi de Zipf, la taille des villes est raisonnablement décrite par un échantillonnage selon une loi à queue lourde, typiquement une loi de puissance avec un faible exposant. Je me suis intéressé à la meilleure façon de tenir compte de cette hétérogénéité dans une stratégie simplifiée d'isolement des villes visant à contenir cette propagation spatiale. A l'aide de deux paramètres synthétiques, nous comparons selon les situations de différents pays les mérites respectifs de stratégies basées sur le taux d'incidence par rapport à des stratégies basées sur le nombre absolu des cas. Ce travail en cours de finalisation est le résultat d'une collaboration transnationale avec Cornelia Pokalyuk, Viktor Bezborodov, Tyll Krueger et Piotr Szymanski.