arXiv subjects

Nicolas Fournier

Publications and source records attributed to Nicolas Fournier.

At least 37 records · Page 2Linked to original sources

arXiv2018

Anomalous diffusion for multi-dimensional critical Kinetic Fokker-Planck equations

We consider a particle moving in $d\geq 2$ dimensions, its velocity being a reversible diffusion process, with identity diffusion coefficient, of which the invariant measure behaves, roughly, like $(1+|v|)^{-β}$ as $|v|\to \infty$, for some constant $β>0$. We prove that for large times, after a suitable rescaling, the position process resembles a Brownian motion if $β\geq 4+d$, a stable process if $β\in [d,4+d)$ and an integrated multi-dimensional generalization of a Bessel process if $β\in (d-2,d)$. The critical cases $β=d$, $β=1+d$ and $β=4+d$ require special rescalings.

math.PR↗

arXiv2018

One dimensional critical Kinetic Fokker-Planck equations, Bessel and stable processes

We consider a particle moving in one dimension, its velocity being a reversible diffusion process, with constant diffusion coefficient, of which the invariant measure behaves like $(1+|v|)^{-β}$ for some $β>0$. We prove that, under a suitable rescaling, the position process resembles a Brownian motion if $β\geq 5$, a stable process if $β\in [1,5)$ and an integrated symmetric Bessel process if $β\in (0,1)$. The critical cases $β=1$ and $β=5$ require special rescaling. We recover some results of G.Lebeau and M.Puel [LP17], P.Cattiaux, E.Nasreddine and M. Puel [CNP16], and E.Barkai, E.Aghion and D. A. Kesslerwith [BAK14] with an alternative approach.

math.PR↗

arXiv2018

On Monte-Carlo tree search for deterministic games with alternate moves and complete information

We consider a deterministic game with alternate moves and complete information, of which the issue is always the victory of one of the two opponents. We assume that this game is the realization of a random model enjoying some independence properties. We consider algorithms in the spirit of Monte-Carlo Tree Search, to estimate at best the minimax value of a given position: it consists in simulating, successively, $n$ well-chosen matches, starting from this position. We build an algorithm, which is optimal, step by step, in some sense: once the $n$ first matches are simulated, the algorithm decides from the statistics furnished by the $n$ first matches (and the a priori we have on the game) how to simulate the $(n+1)$-th match in such a way that the increase of information concerning the minimax value of the position under study is maximal. This algorithm is remarkably quick. We prove that our step by step optimal algorithm is not globally optimal and that it always converges in a finite number of steps, even if the a priori we have on the game is completely irrelevant. We finally test our algorithm, against MCTS, on Pearl's game and, with a very simple and universal a priori, on the games Connect Four and some variants. The numerical results are rather disappointing. We however exhibit some situations in which our algorithm seems efficient.

math.PR↗

arXiv2017

A recursive algorithm and a series expansion related to the homogeneous Boltzmann equation for hard potentials with angular cutoff

We consider the spatially homogeneous Boltzmann equation for hard potentials with angular cutoff. This equation has a unique conservative weak solution $(f_t)_{t\geq 0}$, once the initial condition $f_0$ with finite mass and energy is fixed. Taking advantage of the energy conservation, we propose a recursive algorithm that produces a $(0,\infty)\times\mathbb{R}^3$ random variable $(M_t,V_t)$ such that $E[M_t {\bf 1}_{\{V_t \in \cdot\}}]=f_t$. We also write down a series expansion of $f_t$. Although both the algorithm and the series expansion might be theoretically interesting in that they explicitly express $f_t$ in terms of $f_0$, we believe that the algorithm is not very efficient in practice and that the series expansion is rather intractable. This is a tedious extension to non-Maxwellian molecules of Wild's sum and of its interpretation by McKean.

math.AP↗

arXiv2016

On the equivalence between some jumping SDEs with rough coefficients and some non-local PDEs

We study some jumping SDE and the corresponding Fokker-Planck (or Kolmogorov forward) equation, which is a non-local PDE. We assume only some measurability and growth conditions on the coefficients. We prove that for any weak solution $(f_t)_{t\in [0,T]}$ of the PDE, there exists a weak solution to the SDE of which the time marginals are given by $(f_t)_{t\in[0,T]}$. As a corollary, we deduce that for any given initial condition, existence for the PDE is equivalent to weak existence for the SDE and uniqueness in law for the SDE implies uniqueness for the PDE. This extends some ideas of Figalli [5] concerning continuous SDEs and local PDEs.

math.PR↗

arXiv2016

On the Kozachenko-Leonenko entropy estimator

We study in details the bias and variance of the entropy estimator proposed by Kozachenko and Leonenko for a large class of densities on $\mathbb{R}^d$. We then use the work of Bickel and Breiman to prove a central limit theorem in dimensions $1$ and $2$. In higher dimensions, we provide a development of the bias in terms of powers of $N^{-2/d}$. This allows us to use a Richardson extrapolation to build, in any dimension, an estimator satisfying a central limit theorem and for which we can give some some explicit (asymptotic) confidence intervals.

math.ST↗

arXiv2016

Statistical inference versus mean field limit for Hawkes processes

We consider a population of $N$ individuals, of which we observe the number of actions as time evolves. For each couple of individuals $(i,j)$, $j$ may or not influence $i$, which we model by i.i.d. Bernoulli$(p)$-random variables, for some unknown parameter $p\in (0,1]$. Each individual acts autonomously at some unknown rate $μ>0$ and acts by mimetism at some rate depending on the number of recent actions of the individuals which influence him, the age of these actions being taken into account through an unknown function $φ$ (roughly, decreasing and with fast decay). The goal of this paper is to estimate $p$, which is the main charateristic of the graph of interactions, in the asymptotic $N\to\infty$, $t\to\infty$. The main issue is that the mean field limit (as $N \to \infty$) of this model is unidentifiable, in that it only depends on the parameters $μ$ and $pφ$. Fortunately, this mean field limit is not valid for large times. We distinguish the subcritical case, where, roughly, the mean number $m_t$ of actions per individual increases linearly and the supercritical case, where $m_t$ increases exponentially. Although the nuisance parameter $φ$ is non-parametric, we are able, in both cases, to estimate $p$ without estimating $φ$ in a nonparametric way, with a precision of order $N^{-1/2}+N^{1/2}m_t^{-1}$, up to some arbitrarily small loss. We explain, using a Gaussian toy model, the reason why this rate of convergence might be (almost) optimal.

math.ST↗

arXiv2015

From a Kac-like particle system to the Landau equation for hard potentials and Maxwell molecules

We prove a quantitative result of convergence of a conservative stochastic particle system to the solution of the homogeneous Landau equation for hard potentials. There are two main difficulties: (i) the known stability results for this class of Landau equations concern regular solutions and seem difficult to extend to study the rate of convergence of some empirical measures; (ii) the conservativeness of the particle system is an obstacle for (approximate) independence. To overcome (i), we prove a new stability result for the Landau equation for hard potentials concerning very general measure solutions. Due to (ii), we have to couple, our particle system with some non independent nonlinear processes, of which the law solves, in some sense, the Landau equation. We then prove that these nonlinear processes are not so far from being independent. Using finally some ideas of Rousset [25], we show that in the case of Maxwell molecules, the convergence of the particle system is uniform in time.

math.PR↗

arXiv2015

Stochastic particle approximation of the Keller-Segel equation and two-dimensional generalization of Bessel processes

The Keller-Segel partial differential equation is a two-dimensional model for chemotaxis. When the total mass of the initial density is one, it is known to exhibit blow-up in finite time as soon as the sensitivity $χ$ of bacteria to the chemo-attractant is larger than $8π$. We investigate its approximation by a system of $N$ two-dimensional Brownian particles interacting through a singular attractive kernel in the drift term. In the very subcritical case $χ\textless{}2π$, the diffusion strongly dominates this singular drift: we obtain existence for the particle system and prove that its flow of empirical measures converges, as $N\to\infty$ and up to extraction of a subsequence, to a weak solution of the Keller-Segel equation. We also show that for any $N\ge 2$ and any value of $χ\textgreater{}0$, pairs of particles do collide with positive probability: the singularity of the drift is indeed visited. Nevertheless, when $χ\textless{}2πN$, it is possible to control the drift and obtain existence of the particle system until the first time when at least three particles collide. We check that this time is a.s. infinite, so that global existence holds for the particle system, if and only if $χ\leq 8π(N-2)/(N-1)$. Finally, we remark that in the system with $N=2$ particles, the difference between the two positions provides a natural two-dimensional generalization of Bessel processes, which we study in details.

math.PR↗

arXiv2015

On a toy model of interacting neurons

We continue the study of a stochastic system of interacting neurons introduced in De Masi-Galves-Löcherbach-Presutti (2014). The system consists of N neurons, each spiking randomly with rate depending on its membrane potential. At its spiking time, the neuron potential is reset to 0 and all other neurons receive an additional amount 1/N of potential. Moreover, electrical synapses induce a deterministic drift of the system towards its center of mass. We prove propagation of chaos of the system, as N tends to infinity, to a limit nonlinear jumping stochastic differential equation. We consequently improve on the results of De Masi-Galves-Löcherbach-Presutti (2014), since (i) we remove the compact support condition on the initial datum, (ii) we get a rate of convergence in $1/\sqrt N$. Finally, we study the limit equation: we describe the shape of its time-marginals, we prove the existence of a unique non-trivial invariant distribution, we show that the trivial invariant distribution is not attractive, and in a special case, we establish the convergence to equilibrium.

math.PR↗

arXiv2015

Finiteness of entropy for the homogeneous Boltzmann equation with measure initial condition

We consider the $3D$ spatially homogeneous Boltzmann equation for (true) hard and moderately soft potentials. We assume that the initial condition is a probability measure with finite energy and is not a Dirac mass. For hard potentials, we prove that any reasonable weak solution immediately belongs to some Besov space. For moderately soft potentials, we assume additionally that the initial condition has a moment of sufficiently high order ($8$ is enough) and prove the existence of a solution that immediately belongs to some Besov space. The considered solutions thus instantaneously become functions with a finite entropy. We also prove that in any case, any weak solution is immediately supported by ${\mathbb {R}}^3$.

math-ph↗

arXiv2015

Propagation of chaos for the Landau equation with moderately soft potentials

We consider the 3D Landau equation for moderately soft potentials ($γ\in(-2,0)$ with the usual notation) as well as a stochastic system of $N$ particles approximating it. We first establish some strong/weak stability estimates for the Landau equation, which are satisfying only when $γ\in [-1,0)$. We next prove, under some appropriate conditions on the initial data, the so-called propagation of molecular chaos, i.e. that the empirical measure of the particle system converges to the unique solution of the Landau equation. The main difficulty is the presence of a singularity in the equation. When $γ\in (-1,0)$, the strong-weak uniqueness estimate allows us to use a coupling argument and to obtain a rate of convergence. When $γ\in (-2,-1]$, we use the classical martingale method introduced by McKean. To control the singularity, we have to take advantage of the regularity provided by the entropy dissipation. Unfortunately, this dissipation is too weak for some (very rare) aligned configurations. We thus introduce a perturbed system with an additional noise, show the propagation of chaos for that perturbed system and finally prove that the additional noise is almost never used in the limit.

math.PR↗

arXiv2014

Rate of convergence of the Nanbu particle system for hard potentials

We consider the (numerically motivated) Nanbu stochastic particle system associated to the spatially homogeneous Boltzmann equation for true hard potentials. We establish a rate of propagation of chaos of the particle system to the unique solution of the Boltzmann equation. More precisely, we estimate the expectation of the squared Wasserstein distance with quadratic cost between the empirical measure of the particle system and the solution. The rate we obtain is almost optimal as a function of the number of particles but is not uniform in time.

math.AP↗

arXiv2014

High dimensional Hawkes processes

We generalise the construction of multivariate Hawkes processes to a possibly infinite network of counting processes on a directed graph $\mathbb G$. The process is constructed as the solution to a system of Poisson driven stochastic differential equations, for which we prove pathwise existence and uniqueness under some reasonable conditions. We next investigate how to approximate a standard $N$-dimensional Hawkes process by a simple inhomogeneous Poisson process in the mean-field framework where each pair of individuals interact in the same way, in the limit $N \rightarrow \infty$. In the so-called linear case for the interaction, we further investigate the large time behaviour of the process. We study in particular the stability of the central limit theorem when exchanging the limits $N, T\rightarrow \infty$ and exhibit different possible behaviours. We finally consider the case $\mathbb G = \mathbb Z^d$ with nearest neighbour interactions. In the linear case, we prove some (large time) laws of large numbers and exhibit different behaviours, reminiscent of the infinite setting. Finally we study the propagation of a {\it single impulsion} started at a given point of $\zz^d$ at time $0$. We compute the probability of extinction of such an impulsion and, in some particular cases, we can accurately describe how it propagates to the whole space.

math.PR↗

arXiv2013

On the rate of convergence in Wasserstein distance of the empirical measure

Let $μ_N$ be the empirical measure associated to a $N$-sample of a given probability distribution $μ$ on $\mathbb{R}^d$. We are interested in the rate of convergence of $μ_N$ to $μ$, when measured in the Wasserstein distance of order $p>0$. We provide some satisfying non-asymptotic $L^p$-bounds and concentration inequalities, for any values of $p>0$ and $d\geq 1$. We extend also the non asymptotic $L^p$-bounds to stationary $ρ$-mixing sequences, Markov chains, and to some interacting particle systems.

math.PR↗

arXiv2012

Propagation of chaos for the 2D viscous vortex model

We consider a stochastic system of $N$ particles, usually called vortices in that setting, approximating the 2D Navier-Stokes equation written in vorticity. Assuming that the initial distribution of the position and circulation of the vortices has finite (partial) entropy and a finite moment of positive order, we show that the empirical measure of the particle system converges in law to the unique (under suitable a priori estimates) solution of the 2D Navier-Stokes equation. We actually prove a slightly stronger result : the propagation of chaos of the stochastic paths towards the solution of the expected nonlinear stochastic differential equation. Moreover, the convergence holds in a strong sense, usually called entropic (there is no loss of entropy in the limit). The result holds without restriction (but positivity) on the viscosity parameter. The main difficulty is the presence of the singular Biot-Savart kernel in the equation. To overcome this problem, we use the dissipation of entropy which provides some (uniform in $N$) bound on the Fisher information of the particle system, and then use extensively that bound together with classical and new properties of the Fisher information.

math.AP↗

arXiv2012

Mean-field forest-fire models and pruning of random trees

We consider a family of discrete coagulation-fragmentation equations closely related to the one-dimensional forest-fire model of statistical mechanics: each pair of particles with masses $i,j \in \nn$ merge together at rate 2 to produce a single particle with mass $i+j$, and each particle with mass $i$ breaks into $i$ particles with mass 1 at rate $(i-1)/n$. The (large) parameter $n$ controls the rate of ignition and there is also an acceleration factor (depending on the total number of particles) in front of the coagulation term. We prove that for each $n\in \nn$, such a model has a unique equilibrium state and study in details the asymptotics of this equilibrium as $n\to \infty$: (I) the distribution of the mass of a typical particle goes to the law of the number of leaves of a critical binary Galton-Watson tree, (II) the distribution of the mass of a typical size-biased particle converges, after rescaling, to a limit profile, which we write explicitly in terms of the zeroes of the Airy function and its derivative. We also indicate how to simulate perfectly a typical particle and a size-biased typical particle, which allows us to give some probabilistic interpretations of the above results in terms of pruned Galton-Watson trees and pruned continuum random trees.

math.PR↗

arXiv2011

Asymptotic of grazing collisions and particle approximation for the Kac equation without cutoff

The subject of this article is the Kac equation without cutoff. We first show that in the asymptotic of grazing collisions, the Kac equation can be approximated by a Fokker-Planck equation. The convergence is uniform in time and we give an explicit rate of convergence. Next, we replace the small collisions by a small diffusion term in order to approximate the solution of the Kac equation and study the resulting error. We finally build a system of stochastic particles undergoing collisions and diffusion, that we can easily simulate, which approximates the solution of the Kac equation without cutoff. We give some estimates on the rate of convergence.

math-ph↗