Searcharxiv⌕ Search

arXiv subjects

Guillaume Poly

Publications and source records attributed to Guillaume Poly.

At least 19 recordsLinked to original sources
arXiv2026

Global universality of the expected number of zeros of non-analytic random signals

We study the asymptotics as $n$ goes to infinity of $\mathbb E\left[\mathcal{N}(S_n,[0,2\pi])\right]$, the expected number of zeros in $[0, 2\pi]$ of a random periodic signal $S_n$ of the form \[ S_n(t)=\sum_{k=1}^{n}a_k f(kt), \] where $f$ is a non-analytic $2\pi-$periodic function and the coefficients $(a_k)$ are i.i.d. random variables, centered with unit variance. We show in particular that if $a_1$ admits a finite third moment and if the function $f$ is piecewise polynomials and of class $\mathcal C^{7}$, then we have the following universal asymptotics, independent of the particular law of the coefficients $(a_k)$ \[ \lim_{n \to +\infty}\frac{\Esp\left[\mathcal{N}(S_n,[0,2\pi])\right]}{n}= \frac{2}{\sqrt{3}}\sqrt{\frac{\|f'\|_{L^2([0,2\pi])}}{\|f\|_{L^2([0,2\pi])}}}. \] This result thus extends in expectation and at the scale of the whole period $[0,2\pi]$ the local universality property established {in [Angst-Poly, IMRN, 2019]}, in distribution and in shrinking intervals of size $1/n$. Moreover, it generalizes to a non-analytic context the global universality results obtained in the more classical frameworks of random trigonometric polynomials or random analytic functions. Our approach combines a new almost sure Central Limit Theorem \`a la Salem--Zygmund for the function $S_n$ when evaluated at a uniform random point in $[0, 2\pi]$, and as well as suitable uniform integrability and anti-concentration estimates.

math.PR↗
arXiv2026

Breuer-Major-Donsker invariance principle

We prove a Breuer-Major-type Donsker's invariance principle for stationary Gaussian sequences under the natural \emph{finite-variance assumption} on the test function. This result, which we call the \emph{Breuer-Major-Donsker} principle, or simply the \emph{BMD principle}, removes the additional moment assumption imposed in the functional Breuer-Major theorem of Nourdin and Nualart (\emph{Probab. Theory Related Fields}, 2020). Our method does not rely on the Malliavin-calculus estimates used by Nourdin and Nualart, in particular Meyer's inequality. Instead, it is based on a predictable-martingale decomposition of the partial-sum process, which is of independent interest. We also make systematic use of \emph{non-determinism}, a central notion in Gaussian prediction theory. In the non-deterministic case, the martingale part is handled by the martingale functional central limit theorem, while the predictable remainder gains integrability above order two through Ornstein-Uhlenbeck smoothing. In the deterministic case, the martingale part vanishes, and the smoothing mechanism is no longer available along the full sequence. Nevertheless, under an additional mild assumption on the covariance function, a suitable decimation recovers non-determinism and reduces the proof to the non-deterministic case.

math.PR↗
arXiv2026

Convergence of higher derivatives of random polynomials with independent roots

Let $\mu$ be a probability measure on $\mathbb C$, and let $P_n$ be the random polynomial whose zeros are sampled independently from $\mu$. We study the asymptotic distribution of zeros of high-order derivatives of $P_n$. We show that, for large classes of measures $\mu$, the empirical distribution of zeros of the $k$-th derivative converges back to $\mu$ for all derivative orders $k=o(n/\log n)$. This includes all discrete measures and a broad family of measures satisfying a mild dimension-nondegeneracy condition. We further establish a robustness result showing that, for arbitrary $\mu$, even after adding a vanishing proportion of roots drawn from a dimension-nondegenerate perturbation, the derivative zero measures still converge back to $\mu$. These results break the previously known logarithmic barrier on the order of differentiation and demonstrate that the limiting root distribution is preserved under differentiation of order growing nearly linearly with the degree.

math.PR↗
arXiv2025

Edgeworth expansion on Wiener chaos

Consider $F$ an element of the $p$-th Wiener chaos $\WW_p$, and denote by $\prob_F$ its law. For a positive integer $m$, let $\boldsymbol{\gamma}_{F,m}$ be the Radon measure with density $x \mapsto \frac{e^{-x^2/2}}{\sqrt{2\pi}} \left(1 + \sum_{k=3}^{4m-1} \frac{\E[H_k(F)]}{k!}\, H_k(x)\right)$, where $H_k$ is the $k$-th Hermite polynomial. The main goal of this article is to prove that the total variation distance between $\prob_F$ and $\boldsymbol{\gamma}_{F,m}$ is of order $\Var(\Gamma(F,F))^{({m+1})/{2}}$, where $\Gamma(F,F)$ denotes the carr\'e-du-champ operator of $F$. The variance of $\Gamma(F,F)$ is known to govern Gaussian fluctuations and can be bounded from above by $\kappa_4(F)$, the fourth cumulant of $F$, as established in the seminal work \cite{NP2009a}. Our result thus provides a genuine Edgeworth expansion in the setting of central convergence on Wiener chaoses. In this context, the quantity $\Var(\Gamma(F,F))$ plays the role of the small parameter that governs the accuracy of the approximation, in the same way that $1/\sqrt{n}$ does in the classical central limit theorem. To the best of our knowledge, our work is the first to establish Edgeworth expansions for Wiener chaoses in full generality and at arbitrary order, together with explicit remainder bounds that systematically improve with the order of the expansion--exactly as one would expect from an Edgeworth approximation. Our results apply verbatim to every situation where a central limit theorem is available for chaos elements, since no structural assumption is required beyond belonging to a fixed Wiener chaos. As a byproduct, we recover the celebrated optimal fourth moment theorem from \cite{NP2015} by combining the expansions at the first and second orders, with sharper quantitative bounds. Previous works on Edgeworth expansions for Wiener chaoses were essentially restricted to the first order.

math.PR↗
arXiv2025

Stein's method for Fr\'echet approximation: a regularly varying functions approach

We develop a variant of Stein's method of comparison of generators to bound the Kolmogorov, total variation, and Wasserstein-1 distances between distributions on the real line. Our discrepancy is expressed in terms of the ratio of reverse hazard rates; it therefore remains tractable even when density derivatives are intractable. Our main application concerns the approximation of normalized extremes by Fr\'echet laws. In this setting, the new discrepancy provides a quantitative measure of distributional proximity in terms of the average regular variation at infinity of the underlying cumulative distribution function. We illustrate the approach through explicit computations for maxima of Pareto, Cauchy, and Burr~XII distributions. Our new discrepancy also opens the way to statistical applications which we outline.

math.PR↗
arXiv2024

Limit distributions for polynomials with independent and identically distributed entries

We characterize the limiting distributions of random variables of the form $P_n\left( (X_i)_{i \ge 1} \right)$, where: (i) $(P_n)_{n \ge 1}$ is a sequence of multivariate polynomials, each potentially involving countably many variables; (ii) there exists a constant $D \ge 1$ such that for all $n \ge 1$, the degree of $P_n$ is bounded above by $D$; (iii) $(X_i)_{i \ge 1}$ is a sequence of independent and identically distributed random variables, each with zero mean, unit variance, and finite moments of all orders. More specifically, we prove that the limiting distributions of these random variables can always be represented as the law of $P_\infty\left( (X_i, G_i)_{i \ge 1} \right)$, where $P_\infty$ is a polynomial of degree at most $D$ (potentially involving countably many variables), and $(G_i)_{i \ge 1}$ is a sequence of independent standard Gaussian random variables, which is independent of $(X_i)_{i \ge 1}$. We solve this problem in full generality, addressing both Gaussian and non-Gaussian inputs, and with no extra assumption on the coefficients of the polynomials. In the Gaussian case, our proof builds upon several original tools of independent interest, including a new criterion for central convergence based on the concept of maximal directional influence. Beyond asymptotic normality, this novel notion also enables us to derive quantitative bounds on the degree of the polynomial representing the limiting law. We further develop techniques regarding asymptotic independence and dimensional reduction. To conclude for polynomials with non-Gaussian inputs, we combine our findings in the Gaussian case with invariance principles.

math.PR↗
arXiv2024

Roots of random trigonometric polynomials with general dependent coefficients

We consider random trigonometric polynomials with general dependent coefficients. We show that under mild hypotheses on the structure of dependence, the asymptotics as the degree goes to infinity of the expected number of real zeros coincides with the independent case. To the best of our knowledge, this universality result is the first obtained in a non-Gaussian dependent context. Our proof highlights the robustness of real zeros, even in the presence of dependencies. These findings bring the behavior of random polynomials closer to real-world models, where dependencies between coefficients are common.

math.PR↗
arXiv2024

Sharp total variation rates of convergence for fluctuations of linear statistics of $\beta$-ensembles

In this article, we revisit the question of fluctuations of linear statistics of beta ensembles in the single cut and non-critical regime for general potentials $V$ under mild regularity and growth assumptions. Our main objective is to establish sharp quantitative Central Limit Theorems (CLT) for strong distances, such as the total variation distance, which to the best of our knowledge, is new for general potentials, even qualitatively. Namely, setting $\mu_V$ the equilibrium measure, for a test function $\xi \in \mathscr{C}^{14}$, we establish the convergence in total variation of $X_n=\sum_{i=1}^n \xi(\lambda_i)-n\langle \xi,\mu_V\rangle$ to an explicit Gaussian variable at the sharp speed $1/n$. Under the same assumptions, we also establish multivariate CLTs for vectors of linear statistics in $p-$Wasserstein distances for any $p\ge 1$, with the optimal rate $1/n$, a result which already in dimension one sharpens the speed of convergence established in the recent contribution [26] as well as the required regularity on the test functions. A second objective of this paper, in a more qualitative direction, is to establish the so-called super-convergence of linear statistics, that is to say the convergence of all derivatives of the densities of $X_n$ uniformly on $\mathbb{R}$, provided that $\xi\in\mathscr{C}^\infty(\mathbb{R})$ and is not too degenerated in some sense.

math.PR↗
arXiv2023

Derivatives of entropy and the MMSE conjecture

We investigate the entropy $H(\mu,t)$ of a probability measure $\mu$ along the heat flow and more precisely we seek for closed algebraic representations of its derivatives. Provided that $\mu$ admits moments of any order, it is indeed proved in [Guo et al., 2010] that $t\mapsto H(\mu,t)$ is smooth, and in [Ledoux, 2016] that its derivatives at zero can be expressed into multivariate polynomials evaluated in the moments (or cumulants) of $\mu$. In the seminal contribution \cite{Led}, these algebraic expressions are derived through $\Gamma$-calculus techniques which provide implicit recursive formulas for these polynomials. Our main contribution consists in a fine combinatorial analysis of these inductive relations and for the first time to derive closed formulas for the leading coefficients of these polynomials expressions. Building upon these explicit formulas we revisit the so-called "MMSE conjecture" from [Guo et al., 2010] which asserts that two distributions on the real line with the same entropy along the heat flow must coincide up to translation and symmetry. Our approach enables us to provide new conditions on the source distributions ensuring that the MMSE conjecture holds and to refine several criteria proved in [Ledoux, 2016]. As illustrating examples, our findings cover the cases of uniform and Rademacher distributions, for which previous results in the literature were inapplicable.

cs.IT↗
arXiv2023

A total variation version of Breuer--Major Central Limit Theorem under $\mathbb{D}^{1,2}$ assumption

In this note, we establish a qualitative total variation version of Breuer--Major Central Limit Theorem for a sequence of the type $\frac{1}{\sqrt{n}} \sum_{1\leq k \leq n} f(X_k)$, where $(X_k)_{k\ge 1}$ is a centered stationary Gaussian process, under the hypothesis that the function $f$ has Hermite rank $d \geq 1$ and belongs to the Malliavin space $\mathbb D^{1,2}$. This result in particular extends the recent works of [NNP21], where a quantitative version of this result was obtained under the assumption that the function $f$ has Hermite rank $d= 2$ and belongs to the Malliavin space $\mathbb D^{1,4}$. We thus weaken the $\mathbb D^{1,4}$ integrability assumption to $\mathbb D^{1,2}$ and remove the restriction on the Hermite rank of the base function. While our method is still based on Malliavin calculus, we exploit a particular instance of Malliavin gradient called the sharp operator, which reduces the desired convergence in total variation to the convergence in distribution of a bidimensional Breuer--Major type sequence.

math.PR↗
arXiv2023

Regularity of laws via Dirichlet forms -- Application to quadratic forms in independent and identically distributed random variables

We study the regularity of the law of a quadratic form $Q(X,X)$, evaluated in a sequence $X = (X_{i})$ of independent and identically distributed random variables, when $X_{1}$ can be expressed as a sufficiently smooth function of a Gaussian field. This setting encompasses a large class of important and frequently used distributions, such as, among others, Gaussian, Beta, for instance uniform, Gamma distributions, or else any polynomial transform of them. Let us present an emblematic application. Take $X = (X_{i})$ a sequence of independent and identically distributed centered random variables, with unit variance, following such distribution. Consider also $(Q_{n})$ a sequence of quadratic forms, with associated symmetric Hilbert--Schmidt operators $(\mathsf{A}^{(n)})$. Assume that $\operatorname{Tr}[ (\mathsf{A}^{(n)})^{2} ] = 1/2$, $\mathsf{A}^{(n)}_{ii} =0$, and the spectral radius of $\mathsf{A}^{(n)}$ tends to $0$. Then, $(Q_{n}(X))$ converges in a strong sense to the standard Gaussian distribution. Namely, all derivatives of the densities, which are well-defined for $n$ sufficiently large, converge uniformly on $\mathbb{R}$ to the corresponding derivatives of the standard Gaussian density. While classical methods, from Malliavin calculus or $\Gamma$-calculus, generally consist in bounding negative moments of the so-called \emph{carr\'e du champ} operator $\Gamma(Q(X),Q(X))$, we provide a new paradigm through a second-order criterion involving the eigenvalues of a Hessian-type matrix related to $Q(X)$. This Hessian is built by iterating twice a tailor-made gradient, the \emph{sharp operator} $\sharp$, obtained via a Gaussian representation of the carr\'e du champ. We believe that this method, recently developed by the authors in the current paper and in their companion paper [AoP 52 n{\deg}3 (2024)] , is of independent interest and could prove useful in other settings.

math.PR↗
arXiv2023

Superconvergence phenomenon in Wiener chaoses

We establish an unexpected phenomenon of strong regularization along normal convergence on Wiener chaoses. For every sequence of chaotic random variables, convergence in law to the Gaussian distribution is upgraded to superconvergence: the regularity of the densities increases along the convergence, and all the derivatives converges uniformly. Our findings strengthen known results regarding modes of convergence for normal approximation on Wiener chaoses. Without additional assumptions, convergence in total variation is established by Nourdin & Peccati, and later on amplified to convergence in relative entropy by Nourdin, Peccati & Swan. Our result is then extended to the multivariate setting, and for polynomial mappings of a Gaussian field provided the projection on the Wiener chaos of maximal degree admits a non-degenerate Gaussian limit. While our findings apply to any context involving polynomials of a Gaussian field, we emphasize applications regarding: improved Carbery-Wright estimates near Gaussianity; normal convergence in entropy and in Fisher information; superconvergence for the spectral moments of GOE; moments bounds for the inverse of strongly correlated Wishart-type matrices; superconvergence in the Breuer-Major Theorem. Our proofs leverage Malliavin's historical idea to establish smoothness of the density via the existence of negative moments of the Malliavin gradient, and we develop a new paradigm to study this problem. We relate the existence of negative moments to spectral quantities associated with the Malliavin Hessian. This link relies on an adequate choice of the Malliavin gradient, which provides a novel decoupling procedure of independent interest. Previous attempts to establish convergence beyond entropy have imposed restrictive assumptions ensuring finiteness of negative moments for the Malliavin derivatives. Our analysis renders these assumptions superfluous.

math.PR↗
arXiv2023

Almost sure behavior of the critical points of random polynomials

Let $(Z_k)_{k\geq 1}$ be a sequence of independent and identically distributed complex random variables with common distribution $\mu$ and let $P_n(X):=\prod_{k=1}^n (X-Z_k)$ the associated random polynomial in $\mathbb C[X]$. In [Kab15], the author established the conjecture stated by Pemantle and Rivin in [PR13] that the empirical measure $\nu_n$ associated with the critical points of $P_n$ converges weakly in probability to the base measure $\mu$. In this note, we establish that the convergence in fact holds in the almost sure sense. Our result positively answers a question raised by Z. Kabluchko and formalized as a conjecture in the recent paper [MV22].

math.PR↗
arXiv2022

A short proof of the strong three dimensional Gaussian product inequality

We prove the strong form of the Gaussian product conjecture in dimension three. Our purely analytical proof simplifies previously known proofs based on combinatorial methods or computer-assisted methods, and allows us to solve the case of any triple of even positive integers which remained open so far.

math.PR↗
arXiv2021

Fluctuations in Salem--Zygmund almost sure central limit theorem

Let us consider i.i.d. random variables $\{a_k,b_k\}_{k \geq 1}$ defined on a common probability space $(\Omega, \mathcal F, \mathbb P)$, following a symmetric Rademacher distribution and the associated random trigonometric polynomials $S_n(\theta)= \frac{1}{\sqrt{n}} \sum_{k=1}^n a_k \cos(k\theta)+b_k \sin(k\theta)$. A seminal result by Salem and Zygmund ensures that $\mathbb{P}-$almost surely, $\forall t\in\mathbb{R}$ \[ \lim_{n \to +\infty} \frac{1}{2\pi}\int_0^{2\pi} e^{i t S_n(\theta)}d\theta=e^{-t^2/2}. \] This result was then further generalized in various directions regarding the coefficients distribution, their dependency structure or else the dimension and the nature of the ambient manifold. To the best of our knowledge, the natural question of the fluctuations in the above limit has not been tackled so far and is precisely the object of this article. Namely, for general i.i.d. symmetric random coefficients having a finite sixth-moment and for a large class of continuous test functions $\phi$ we prove that \[ \sqrt{n}\left(\frac{1}{2\pi}\int_0^{2\pi} \phi(S_n(\theta))d\theta-\int_{\mathbb{R}}\phi(t)\frac{e^{-\frac{t^2}{2}}dt}{\sqrt{2\pi}}\right)\xrightarrow[n\to\infty]{\text{Law}}~\mathcal{N}\left(0,\sigma_{\phi}^2+\frac{c_2(\phi)^2}{2}\left(\mathbb{E}(a_1^4)-3\right)\right). \] Here, the constant $\sigma_{\phi}^2$ is explicit and corresponds to the limit variance in the case of Gaussian coefficients and $c_2(\phi)$ is the coefficient of order $2$ in the decomposition of $\phi$ in the Hermite polynomial basis. Surprisingly, it thus turns out that the fluctuations are not universal since they both involve the kurtosis of the coefficients and the second coefficient of $\phi$ in the Hermite basis.

math.PR↗
arXiv2021

Real zeros of random trigonometric polynomials with dependent coefficients

We further investigate the relations between the large degree asymptotics of the number of real zeros of random trigonometric polynomials with dependent coefficients and the underlying correlation function. We consider trigonometric polynomials of the form \[ f_n(t):= \frac{1}{\sqrt{n}}\sum_{k=1}^{n}a_k \cos(kt)+b_k\sin(kt), ~x\in [0,2\pi], \] where the sequences $(a_k)_{k\geq 1}$ and $(b_k)_{k\geq 1}$ are two independent copies of a stationary Gaussian process centered with variance one and correlation function $\rho$ with associated spectral measure $\mu_{\rho}$. We focus here on the case where $\mu_{\rho}$ is not purely singular and we denote by $\psi_{\rho}$ its density component with respect to the Lebesgue measure $\lambda$. Quite surprisingly, we show that the asymptotics of the number of real zeros $\mathcal{N}(f_n,[0,2\pi])$ of $f_n$ in $[0,2\pi]$ is not related to the decay of the correlation function $\rho$ but instead to the Lebesgue measure of the vanishing locus of $\psi_{\rho}$. Namely, assuming that $\psi_{\rho}$ is $\mathcal{C}^1$ with H\"older derivative on an open set of full measure, one establishes that \[ \lim_{n \to +\infty} \frac{\mathbb E\left[\mathcal{N}(f_n,[0,2\pi])\right]}{n}= \frac{\lambda(\{\psi_{\rho}=0\})}{\pi \sqrt{2}} + \frac{2\pi - \lambda(\{\psi_{\rho}=0\})}{\pi\sqrt{3}}. \] On the other hand, assuming a sole log-integrability condition on $\psi_{\rho}$, which implies that it is positive almost everywhere, we recover the asymptotics of the independent case, i.e. the limit is $\frac{2}{\sqrt{3}}$. Besides, with further assumptions of regularity and existence of negative moment for $\psi_{\rho}$, we moreover show that the above convergence in expectation can be strengthened to an almost sure convergence.

math.PR↗
arXiv2019

Variations on Salem--Zygmund results for random trigonometric polynomials. Application to almost sure nodal asymptotics

On a probability space $(Ω, \mathcal F, \mathbb P)$ we consider two independent sequences $(a_k)_{k \geq 1}$ and $(b_k)_{k \geq 1}$ of i.i.d. random variables that are centered with unit variance and which admit a moment strictly higher than two. We define the associated random trigonometric polynomial \[ f_n(t) :=\frac{1}{\sqrt{n}} \sum_{k=1}^n a_k \cos(kt)+b_k \sin(kt), \quad t \in \mathbb R. \] In their seminal work, for Rademacher coefficients, Salem and Zygmund showed that $\mathbb P$ almost surely: \[ \forall t\in\mathbb R,\quad \frac{1}{2π}\int_{0}^{2π} \exp\left(i t f_n(x)\right) dx \xrightarrow[n\to\infty]~e^{-\frac{t^2}{2}}. \] In other words, if $X$ denotes an independent random variable uniformly distributed on $[0,2π]$, $\mathbb{P}$ almost surely, under the law of $X$, $f_n(X)$ converges in distribution to a standard Gaussian variable. In this paper, we revisit the above result from different perspectives. Namely, i) we establish a possibly sharp convergence rate for some adequate metric via the Stein's method, ii) we prove a functional counterpart of Salem--Zygmund CLT, iii) we extend it to more general distributions for $X$, iv) we also prove that the convergence actually holds in total variation. As an application, in the case where the random coefficients have a symmetric distribution and admit a moment of order $4$, we show that $\mathbb{P}$ almost surely, for any interval $[a,b] \subset [0, 2π]$ \[\frac{\mathcal N(f_n,[a,b])}{n} \xrightarrow[n \to +\infty]{} \frac{(b-a)}{π\sqrt{3}},\] where $\mathcal N(f_n,[a,b])$ denotes the number of real zeros of $f_n$ in the interval $[a,b]$. To the best of our knowledge, such an almost sure result is new in the framework of random trigonometric polynomials, even in the case of Gaussian coefficients.

math.PR↗
arXiv2019

On the zeros of non-analytic random periodic signals

In this paper, we investigate the local universality of the number of zeros of a random periodic signal of the form $S_n(t)=\sum_{k=1}^n a_k f(k t)$, where $f$ is a $2π-$periodic function satisfying weak regularity conditions and where the coefficients $a_k$ are i.i.d. random variables, that are centered with unit variance. In particular, our results hold for continuous piecewise linear functions. We prove that the number of zeros of $S_n(t)$ in a shrinking interval of size $1/n$ converges in law as $n$ goes to infinity to the number of zeros of a Gaussian process whose explicit covariance only depends on the function $f$ and not on the common law of the random coefficients $(a_k)$. As a byproduct, this entails that the point measure of the zeros of $S_n(t)$ converges in law to an explicit limit on the space of locally finite point measures on $\mathbb R$ endowed with the vague topology. The standard tools involving the regularity or even the analyticity of $f$ to establish such kind of universality results are here replaced by some high-dimensional Berry-Esseen bounds recently obtained in [CCK17]. The latter allow us to prove functional CLT's in $C^1$ topology in situations where usual criteria can not be applied due to the lack of regularity.

math.PR↗