Weak approximation for Gaussian processes from renewal processes All authors are supported by the grant PID2021-123733NB-I00 funded by MCIN/AEI/10.13039/501100011033. Salim Boukfal is also supported by the program of predoctoral grants AGAUR-FI (2025 FI-1 00119) Joan Oró from the Department of Resea
Xavier Bardina, Salim Boukfal, Marc Cano, Carles Rovira · 2025
Read Weak approximation for Gaussian processes from renewal processes All authors are supported by the grant PID2021-123733NB-I00 funded by MCIN/AEI/10.13039/501100011033. Salim Boukfal is also supported by the program of predoctoral grants AGAUR-FI (2025 FI-1 00119) Joan Oró from the Department of Resea by Xavier Bardina, Salim Boukfal, Marc Cano, Carles Rovira (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-17280.txt.
Introduction
Xavier Bardina and three co-authors extend a family of approximations that build Brownian motion from renewal processes, random event streams with independent gaps. In earlier work, Bardina and Rovira built such processes converging strongly to Brownian motion. Here the team shows some of those processes can approximate richer Gaussian processes, including fractional Brownian motion, which has long-range memory, and multiple Stratonovich integrals.
The paper gives sufficient conditions on the underlying renewal process for the convergence to hold. Fractional Brownian motion with any Hurst parameter serves as the illustrative example, covering both rougher and smoother behavior than ordinary Brownian motion. Approximations of this kind matter because Gaussian processes are usually defined abstractly, while the renewal constructions are concrete paths one can simulate and manipulate. The results tie a simple random mechanism to a broad class of Gaussian limits.
Catalog record
| CREATOR | Xavier Bardina, Salim Boukfal, Marc Cano, Carles Rovira |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-17280
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Weak approximation for Gaussian processes from renewal processes All authors are supported by the grant PID2021-123733NB-I00 funded by MCIN/AEI/10.13039/501100011033. Salim Boukfal is also supported by the program of predoctoral grants AGAUR-FI (2025 FI-1 00119) Joan Oró from the Department of Research and Universities of the Government of Catalonia and the co-funding of the European Social Fund Plus (ESF+). Marc Cano is also supported by the predoctoral grant PRE2022-103868 funded by MCIN/AEI/ 10.13039/501100011033. Xavier Bardina, Salim Boukfal, Marc Cano, Carles Rovira Abstract In [ 4 ] a family of processes that converge strongly towards Brownian motion, defined from renewal processes, are constructed. In this paper we prove that some of these processes can be utilized to build approximations of Gaussian processes such as fractional Brownian motion or multiple Stratonovich integrals and we provide sufficient conditions on renewal processes to ensure that the convergence holds. An illustrative example of such a Gaussian process is the fractional Brownian motion with any Hurst parameter. Keywords. weak convergence, renewal process, fractional Brownian motion, multiple Stratonovich integrals 1 Introduction Let \{N(t),\,t\geq 0\} a standard Poisson process and, for all n\in\mathbb{N} , consider the process: h_{n}=\left\{h_{n}(t):=\frac{1}{\sqrt{n}}\int_{0}^{nt}(-1)^{N(u)}du,\,t\in[0,T]\right\}. In 1982, Stroock (see [ 10 ] ) established that the processes h_{n} converge in law to a standard Brownian motion. These processes were initially introduced by Kac in 1956 (see [ 7 ] ) to derive a solution to the telegraph equation from a Poisson process. Several extensions of the Kac–Stroock result have been investigated in the literature. These include modifying the processes h_{n} to approximate Gaussian processes other than Brownian motion, establishing stronger forms of convergence, and relaxing the structural assumptions on the approximating processes in order to obtain convergence to Brownian motion in more general settings. An extension in the first direction, that is, modifying the processes h_{n} to obtain approximations of other Gaussian processes, is presented in [ 5 ] by Delgado and Jolis, who consider a class of Gaussian processes encompassing the fractional Brownian motion. For a stochastic process Y with representation: Y=\{Y_{t}=\int_{0}^{1}K(t,r)dW_{r},\,t\in[0,1]\}, where W is a standard Brownian motion and the kernel K:[0,1]\times[0,1]\longrightarrow\mathbb{R} satisfies certain properties, they define the process {Y}_{t}^{n}={\sqrt{n}}\int_{0}^{1}K(t,r)(-1)^{N(nr)}dr. Their result proves that {Y}_{t}^{n} converges in law to Y in the space \mathcal{C}([0,1]) . Also in this direction, in [ 2 ] we find results of convergence to Stratonovich multiple integrals. The second direction of extension concerns proving convergence in a stronger sense than convergence in distribution on the space of continuous functions. Several works obtain such stronger convergence results, in which the realizations of the processes converge almost surely and uniformly on the unit interval to Brownian motion. This is achieved by modifying the processes so that they do not start with a strictly increasing behaviour. The resulting modified processes, commonly referred to as uniform transport processes, are X_{n}(t)=\frac{1}{\sqrt{n}}(-1)^{A}\int_{0}^{nt}(-1)^{N(u)}du, where A\sim\textrm{Bern}\left(\frac{1}{2}\right) is independent of the Poisson process N . Griego, Heath and Ruiz-Moncayo [ 6 ] showed that these processes converge strongly and uniformly on bounded time intervals to Brownian motion. The third direction is weakening the conditions of the approximating processes to find generalizations of the processes (-1)^{N(u)} that also converge to Brownian motion. In [ 3 ] we prove that if instead of using the Poisson process we consider a process with independent increments which characteristic function satisfies some properties we obtain also the convergence towards the law of a Brownian motion. Between the examples of processes that satisfy these conditions there are included the Lévy processes. Finally, in the work of Bardina and Rovira [ 4 ] , the authors introduce a family of processes constructed from a renewal process, which converge almost surely to Brownian motion. The objective of the present paper is to study two results concerning the weak approximation of Gaussian processes arising from these realizations of Brownian motion. In this sense, our contribution extends the framework in two directions: firstly, by approximating Gaussian processes other than Brownian motion, and secondly, by establishing convergence results for more general processes than those originally considered by Kac and Stroock. To achieve this, we follow an approach analogous to that of Delgado and Jolis [ 5 ] for our first main result (Theorem 3.1 ), while the proof of our second result (Theorem 4.1 ) is inspired by the ideas developed in Bardina and Jolis [ 2 ] . To the best of the authors’ knowledge, these are the first weak convergence results obtained from processes constructed using renewal processes available in the literature. They shed light on the conditions required to ensure convergence once the convenient properties of the exponential inter-arrival distribution of the Poisson process are no longer present. The paper is organized as follows. In Section 2, we introduce the renewal processes and the kernels that will be used to construct our approximations. Section 3 is devoted to the convergence towards integrals with respect to a deterministic kernel, while Section 4 addresses the convergence towards multiple integrals. Finally, an appendix is included, where we discuss renewal processes with geometrically distributed inter-arrival times. 2 Renewal process Consider \left\{U_{m}\right\}_{m\geq 1} a sequence of independent random variables which take only nonnegative values. Furthermore, assume that they are identically distributed with \mathbb{P}\left(U_{1}=0\right)<1 and \mathbb{E}\left[\left(U_{1}\right)^{l}\right]<\infty , for l large enough (for our purposes, l\geq 4 suffices). For each k\geq 1 , consider the renewal sequence S_{k}=U_{1}+\cdots+U_{k} and the counting renewal function L(t)=\sum_{k=1}^{\infty}\mathbbm{1}_{[0,t]}(S_{k}). This renewal function counts the number of renewals in [0,t] . Set \left\{\eta_{m}\right\}_{m\geq 0} a sequence of independent identically distributed random variables following a Bernoulli(\frac{1}{2}) distribution and independent of the sequence \left\{U_{m}\right\}_{m\geq 1} . Then, we can construct the following renewal reward process T(t)=\eta_{0}+\sum_{k=1}^{\infty}\eta_{k}\mathbbm{1}_{[0,t]}(S_{k})=\sum_{l=0}^{L(t)}\eta_{l}. Given a strictly positive function \beta , we can define T_{n}(t)=T_{\beta(n)}(t)=T\left(\frac{t}{\beta(n)}\right)=\eta_{0}+\sum_{k=1}^{\infty}\eta_{k}\mathbbm{1}_{\left[0,\frac{t}{\beta(n)}\right]}(S_{k}). Notice that, setting U_{m}^{n}=\beta(n)\times U_{m} , for all m\geq 1 , we have that \beta(n)\times S_{k}=U_{1}^{n}+\cdots+U_{k}^{n}. We will consider the kernels \displaystyle\theta_{n}(x):=\frac{1}{G(n)}(-1)^{T\left(\frac{x}{\beta(n)}\right)}, (2.1) where G(n)=\left(\beta(n)\frac{\mathbb{E}\left[\left(U_{1}\right)^{2}\right]}{\mathbb{E}\left[U_{1}\right]}\right)^{\frac{1}{2}}, with \sum_{n\geq 1}\beta(n)<\infty. Let us recall that, under these hypotheses, in [ 4 ] we prove that the processes x_{n}(t)=\int_{0}^{t}\theta_{n}(u)du (2.2) converge almost surely to the Brownian motion. If, in addition, we assume that U_{1} has a failure rate function bounded away from 0, we can go a step further and show the convergence of processes constructed from the x_{n} such as stochastic integrals or iterated integrals with respect to these (see Theorems 3.1 and 4.1 ). More precisely, we will assume that the cumulative distribution function F of U_{1} satisfies the following hypothesis: (G) F is given by F(t)=1-\exp\left\{-\int_{0}^{t}r(s)ds\right\},\hskip 9.24994ptt\geq 0, (and F(t)=0 otherwise) for some non-negative function r such that \int_{0}^{\infty}r(s)ds=\infty and for which there is a positive constant \lambda>0 such that r(s)\geq\lambda for every s\geq 0 . 3 Convergence to Gaussian processes represented by a stochastic integral of a deterministic kernel Let X=\left\{X_{t},~t\in[0,1]\right\} be a centered Gaussian process with covariance function given by Cov\left(X_{t},X_{s}\right)=\int_{0}^{1}K(t,x)K(s,x)dx, where K\colon[0,1]\times[0,1]\to\mathbb{R} is a function satisfying the following assumptions: (H)\left\{\begin{matrix}[l](i)&&K~\text{is measurable and }K(0,x)=0~\forall x\in[0,1].\\ \\ (ii)&&\text{There exists a continuous increasing function }G\colon[0,1]\to\mathbb{R}\text{ and}\\ &&\text{a constant }\alpha>0~\text{such that for any }0\leq s<t\leq 1,\\ \\ &&\displaystyle\int_{0}^{1}\left(K(t,x)-K(s,x)\right)^{2}dx\leq\left(G(t)-G(s)\right)^{\alpha}.\end{matrix}\right. (3.1) In our first theorem, we establish that the law of a continuous Gaussian process, represented by a stochastic integral of a deterministic kernel with respect to a standard Wiener process, can be weakly approximated by the law