Le modèle IDLA (Internal Diffusion Limited Aggregation) est un modèle de croissance aléatoire sur la grille Z^d introduit dans les années 90 et permettant de modéliser l’évolution d’un bassin de culture de cellules, la croissance de zones urbaines ou encore la propagation d’un fluide visqueux. C’est une suite d’ensembles aléatoires (A_n)_n définie comme suit : A₀ = {0} et, étant donné A_n, on lance une marche aléatoire simple depuis l’origine de Z^d. Le sommet z par lequel la marche sort de l’agrégat A_n est ajouté pour obtenir A_n+1 : A_n+1 = A_n U {z}. Un arbre aléatoire se cache derrière la suite des agrégats (A_n)_n… Afin d’étudier la géométrie de cet arbre, nous avons défini en 2020 un graphe aléatoire auxiliaire, baptisé la forêt dirigée IDLA. Ce nouvel objet possède d’intéressantes propriétés et des conjectures excitantes qui seront abordées dans cet exposé. Travail en collaboration avec N. Chenavier (ULCO) et A. Rousselle (Dijon)

With the emergence of numerical sensors in many aspects of every-day life, there is an increasing need in analyzing high frequency data, which can be seen as discrete observation of functional data.
The presentation will focus on the clustering of such functional data, in order to ease their modeling and understanding. To this end, a novel clustering technique for multivariate functional data is presented.
This method is based on a functional latent mixture model which fits the data in group-specific functional subspaces through a multivariate functional principal component analysis.
In such clustering analysis, the presence of outliers can confuse the notion of cluster.
Consequently, a contaminated version of the previous mixture model is proposed. This model both clusters the multivariate functional data into homogeneous groups and detects outliers. The main advantage of this procedure over its competitors is that it does not require us to specify the proportion of outliers.
Model inference is performed through an Expectation-Conditional Maximization algorithm, and the BIC criterion is used to select the number of clusters. Numerical experiments on simulated data demonstrate the high performance achieved by the inference algorithm. In particular, the proposed model outperforms competitors. Its application on the real data which motivated this study allows us to correctly detect abnormal behaviors.

L’un des principes de l’analyse topologique des données est d’étudier un ensemble de données, représentées sous forme d’un nuage de points, à partir d’outils topologiques. Un concept de base est celui de l’homologie persistante. Cette dernière mesure les naissances et les morts de diverses caractéristiques topologiques, telles que les boucles et les cavités, lorsque l’on fait grossir des boules en chaque point d’un processus de Poisson (on parle de modèle Booléen). Dans cet exposé, nous nous intéressons aux durées de vie extrémales pour de telles caractéristiques. Nous étudions d’abord le cas particulier des cavités et donnons l’ordre de grandeur du maximum (resp. du minimum) de leurs durées de vie. Une approximation poissonienne du nombre d’excédents est également établie. Nous étendons ensuite l’étude à des caractéristiques quelconques pour les complexes simpliciaux de Cech et de Vietoris-Rips. Travail joint avec C. Hirsch.

Je présenterai dans cet exposé des résultats nouveaux sur l’existence et l’unicité de solution pour des EDSRs réfléchies dans des domaines non convexes supposés « faiblement étoilés ». Notons que le cas particulier des EDSRs de générateur nul, à savoir l’espérance conditionnelle pour la filtration brownienne, est déjà un cas d’étude intéressant et permet de définir une notion de moyenne contrainte à prendre ses valeurs dans un ensemble non convexe. En particulier, on établit des résultats d’existence et d’unicité dans un cadre markovien avec une condition terminale et un générateur réguliers, mais également dans un cadre général sous une hypothèse de petitesse sur les paramètres de l’EDSR. C’est un travail en commun avec Jean-François Chassagneux (Université de Paris) et Sergey Nadtochiy (Illinois Institute of Technology).

Reinforced processes are known to provide a stochastic approximation for the quasi-stationary distributions of killed Markov processes. We show how the construction may be adapted to provide a stochastic approximation of the paths of killed Markov processes conditioned on survival. Whilst rigorous results are restricted to time being discrete and the state space finite, the strategy employed should be extendable to a general setting in the future.

Existence and uniqueness of solutions to the stochastic heat equation with multiplicative spatial noise is studied. In the spirit of pathwise regularization by noise, we show that a perturbation by a sufficiently irregular continuous path establish wellposedness of such equations, even when the drift and diffusion coefficients are given as generalized functions or distributions. In addition we prove regularity of the averaged field associated to a Lévy fractional stable motion, and use this as an example of a perturbation regularizing the multiplicative stochastic heat equation.

Joint work With Fabian Harang

Dans cet exposé je présenterai un travail en collaboration avec Oriane Blondel et Cristina Toninelli où nous étudions le modèle FA1f en dimension 1. Il s’agit d’un système de particules en interaction (plus précisément un modèle issu de la physique statistique dit modèle cinétiquement contraint où chaque site met à jour la valeur de son spin si une certaine contrainte locale est satisfaite, ici c’est le fait d’avoir au moins un 0 dans ses voisins). Dans ce travail, nous prouvons, sous certaines conditions, une vitesse linéaire, et des fluctuations gaussiennes, pour le front (i.e. le 0 le plus à gauche lorsque l’on part d’une configuration initiale avec que des 1 à gauche de l’origine et un 0 en l’origine). Ce talk sera l’occasion de présenter les techniques classiques utilisées dans les modèles de croissance aléatoire tels que le processus de contact et de parler de méthode de couplage permettant de passer d’un modèle bien connu a un modèle plus complexe (en particulier non attractif).

Point-cloud datasets are ubiquitous in many science and non-science fields. These data are usually coming along with unique patterns that some algorithms are meant to extract and that are linked with the underlying phenomenon that generated the data.
In this presentation, motivated by cosmological problematics, we will focus on two kinds of spatially structured datasets. First, clustered-type patterns in which the datapoints are separated in the input space into multiple groups. We will show that the unsupervised clustering procedure performed with a Gaussian Mixture Model can be formulated in terms of a statistical physics optimisation problem. This formulation enables the unsupervised extraction of many key information about the dataset itself, like the number of clusters, their size and how they are embedded in space, particularly interesting for high-dimensional input spaces where visualisation is not possible.
On the other hand, we will study spatially continuous datasets assuming as standing on an underlying 1D structure that we aim to learn. To this end, we resort to a regularisation of the Gaussian Mixture Model in which a spatial graph is used as a prior to approximate the underlying 1D structure. The overall graph is efficiently learnt by means of the Expectation-Maximisation algorithm with guaranteed convergence and comes together with the learning of the local width of the structure. We then illustrate applications of the algorithm to model and identify the filamentary pattern drawn by the galaxy distribution of the Universe in cosmological datasets.

In this talk, we will discuss some results on the estimation of the invariant density associated to a multivariate diffusion X = (X_t)_t≥0, assuming that a continuous record of observations (X_t)_0≤t≤T is available. We will see that, when X = (X_t)_t≥0 is the solution of a stochastic differential equation with Levy-type jumps, it is possible to find the parametric convergence rate 1/T in the monodimensional case and log(T)/T when the dimension d is equal to 2. For d ≥ 3 we find the convergence rate (log(T)/T)^γ, where γ is an explicit exponent depending on the dimension d and on β₃, the harmonic mean of the smoothness of the invariant density over the d directions after having removed β₁ and β₂, which are the smallest. Moreover, we obtain a lower bound on the L²-risk for pointwise estimation, with the rate (1/T)^γ. In order to fill the logarithmic gap we consider then X = (X_t)_t≥0 as a solution to a continuous stochastic differential equation. One (surprising) finding is that the convergence rate depends on the fact that β₂ < β₃ or β₂ = β₃. In particular, we show that kernel density estimators achieve the rate (log(T)/T)^γ in the first case and (1/T)^γ in the second. Finally, we prove a minimax lower bound on the L²-risk for the pointwise estimation with the same rates (log(T)/T)^γ or (1/T)^γ, depending on the value of β₂ and β₃.

One of the seemingly innocent reformulations of the terrifying Riemann Hypothesis (RH) is the Nyman-Beurling criterion: The indicator function of (0,1) can be linearly approximated in a L^2 space by dilations of the fractional part function. Randomizing these dilations generates new structures and criteria for RH, regularizing very intricate ones. One other possible nice feature is to consider polynomials instead of Dirichlet polynomials for the approximations. How then are the huge difficulties reallocated? The answers are quite surprising!

The talk will be very accessible, especially for graduate students.
Joint work with F. Alouges and E. Hillion.

Dans cet exposé, on construit, pour tout automate cellulaire probabiliste uni-dimensionnel exponentiellement ergodique et possédant une propriété de taux positifs, un flot CFTP (« coupling from the past ») localement défini. Plusieurs conséquences de cette construction sont discutées. (Travail exposé dans l’article arXiv:2106.07219).

Je montre comment à partir d’une étude approfondie statistique et probabiliste d’une base de données de températures moyennes mondiales, j’ai découvert des comportements climatologiques sans doute très difficilement accessibles par les techniques usuelles utilisées en climatologie.

In a recent paper, Leonid Petrov showed that the up-down chains associated to the Chinese Restaurant Process (CRP) have a scaling limit – namely, a two-parameter family of diffusions that extend the one-parameter infinitely-many-neutral-alleles diffusions of Ethier and Kurtz. There has since been considerable interest in constructing ordered analogues of Petrov’s diffusions, and it is conjectured that an ordered analogue of the up-down chains will give rise to such an object. In this talk, I’ll discuss my resolution of this conjecture (joint with Douglas Rizzolo). Our approach is mainly inspired by Petrov’s work, and involves using quasisymmetric functions to describe the transition operators.

This presentation is a summary of my PhD work. I focus on the topic of hypothesis testing, extensively studied in statistics and theoretical computer science.

I start with presenting the classical identity testing problem, in which an independent sample set X ~ q is given and one would like to determine whether q=p for some fixed known p. This problem is very related to that of estimating a distribution from a given sample set. The study of testing is relevant, because for the same fixed sample size, it is possible to test against a distribution up to a smaller separation distance than what is possible in estimation. This will give me the opportunity to describe the minimax framework which proves the theoretical optimality of statistical methods in the worst case.

I will refine the study of minimax identity testing by adding a local differential privacy condition and the interest will be in the quantitative effect of ensuring privacy. The presentation will largely be on the topic of privacy, because it bears similarities with ensuring fairness conditions.

We will also shortly consider the neighboring problem of closeness testing, where the goal remains to determine whether p=q, but only an independent sample set Y ~ p is given instead of p directly. In this context, we will go beyond a simple worst-case analysis and develop instance optimal results instead. This will highlight the interplay between one-sample testing and two-sample testing, the latter being a harder problem.

Télécharger le résumé (pdf)

Differentiation is the process whereby a cell acquires a specific phenotype, by differential gene expression as a function of time. This is thought to result from the dynamical functioning of an underlying Gene Regulatory Network (GRN). The precise path from the stochastic GRN behavior to the resulting cell state is still an open question. In this presentation, we detail a methodology to reduce a mechanistic model characterizing the evolution of a cell by a system of piecewise deterministic Markov processes (PDMP), to a discrete coarse-grained model on a limited number of cell types, defined as the basins of attraction of the deterministic limit. The transitions between the basins in the weak noise limit can be determined by the unique solution of an Hamilton-Jacobi equation under a particular constraint, which corresponds to the rate function associated to a Large Deviations Principle for the PDMP. We develop a numerical method for approximating the coarse-grained model parameters, and show its accuracy for a toggle-switch network. We deduce from the reduced model an analytical approximation of the stationary distribution of the PDMP system, which appears as a Beta mixture.

Les modèles d’urnes sont utilisés dans de nombreuses applications et sont un exemple fondamental de processus stochastiques de renforcement. En partant de ces modèles, nous nous intéresserons à plusieurs familles de systèmes (finis) de processus de renforcement. Différents résultats sur le comportement collectif en temps long seront présentés. La présence/absence de synchronisation sera discutée, ainsi que les vitesses de convergence en fonction de différents régimes de paramètres. Cet exposé se fonde sur des travaux en collaboration avec I. Crimaldi, P. Dai Pra, I. Minelli et M. Mirebrahimi.

Quasi-stationary distributions can be seen as the first eigenvector associated with the generator of the stochastic differential equation at hand, on a domain with Dirichlet boundary conditions (which corresponds to absorbing boundary conditions at the level of the underlying stochastic processes). Many results on the quasi-stationary distribution hold for non degenerate stochastic dynamics, whose associated generator is elliptic. The case of degenerate dynamics is less clear. In this work, together with T. Lelièvre and J. Reygner (Ecole des Ponts, France) we generalize well-known results on the probabilistic representation of solutions to parabolic equations on bounded domains to the so-called kinetic Fokker-Planck equation on bounded domains in positions, with absorbing boundary conditions. Furthermore, a Harnack inequality, as well as a maximum principle, is provided for solutions to this kinetic Fokker-Planck equation, as well as the existence of a smooth transition density for the associated absorbed Langevin dynamics. The continuity of this transition density at the boundary is studied as well as the compactness, in various functional spaces, of the associated semigroup. This work is a cornerstone to prove the consistency of some algorithms used to simulate metastable trajectories of the Langevin dynamics, for example the Parallel Replica algorithm.

Considérons la percolation de premier passage dans le réseau renormalisé Z^d/n pour d>=2 : à chaque arête e, on associe une capacité aléatoire c(e)>=0 de telle sorte que la famille (c(e))_e soit indépendante et identiquement distribuée selon une loi G. On peut interpréter cette capacité comme un débit maximal, i.e., la quantité maximale d’eau pouvant traverser l’arête par unité de temps. Considérons un domaine borné et connecté Ω de R^d et deux ensembles disjoints du bord de Ω : un part lequel l’eau peut entrer (la source) et un part lequel l’eau peut sortir (le puits). Nous nous intéressons au flot maximal : la quantité maximale d’eau pouvant entrer dans Ω par unité de temps. Un courant est une fonction sur les arêtes qui décrit la façon dont l’eau circule dans Ω. Dans cet exposé, nous présenterons un principe de grande déviation pour les courants et nous en déduirons par un principe de contraction un principe de grande déviation pour le flot maximal dans Ω.
Travail en collaboration avec Marie Théret.

A growth-fragmentation is a stochastic process representing cells with continuously growing mass, which experience sudden splitting events. Growth-fragmentations are used to model cell division and protein polymerisation in biophysics. It is interesting to ask whether these processes converge toward an equilibrium, in which the number of cells is growing exponentially and the distribution of cell sizes approaches some fixed asymptotic profile. In this work, we study a process in which the growth and splitting of an individual cell is largely independent of its mass, with the exception that the mass is bounded above, so it cannot exceed a given constant. We give precise conditions to ensure that, almost surely, the process exhibits this equilibrium behaviour, and express the asymptotic profile in terms of an underlying Lévy process.

This is joint work with Emma Horton (Inria Bordeaux).