Loading [MathJax]/jax/output/SVG/jax.js
Research article

On the central limit theorem for the elephant random walk with gradually increasing memory and random step size

  • Received: 28 March 2024 Revised: 11 May 2024 Accepted: 13 May 2024 Published: 24 May 2024
  • MSC : 60F05, 60F25, 60G42, 60G50

  • In this paper, we investigate an extended version of the elephant random walk model. Unlike the traditional approach where step sizes remain constant, our model introduces a novel feature: step sizes are generated as a sequence of positive independent and identically distributed random variables, and the step of the walker at time n+1 depends only on the steps of the walker between times 1,...,mn, where (mn)n1 is a sequence of positive integers growing to infinity as n goes to infinity. Our main results deal with the validity of the central limit theorem for this new variation of the standard ERW model introduced by Schütz and Trimper in 2004.

    Citation: Rafik Aguech. On the central limit theorem for the elephant random walk with gradually increasing memory and random step size[J]. AIMS Mathematics, 2024, 9(7): 17784-17794. doi: 10.3934/math.2024865

    Related Papers:

    [1] Chao Wei . Parameter estimation for partially observed stochastic differential equations driven by fractional Brownian motion. AIMS Mathematics, 2022, 7(7): 12952-12961. doi: 10.3934/math.2022717
    [2] Frithjof Lutscher, Thomas Hillen . Correlated random walks in heterogeneous landscapes: Derivation, homogenization, and invasion fronts. AIMS Mathematics, 2021, 6(8): 8920-8948. doi: 10.3934/math.2021518
    [3] Najmeddine Attia . Some remarks on recursive sequence of fibonacci type. AIMS Mathematics, 2024, 9(9): 25834-25848. doi: 10.3934/math.20241262
    [4] Mohamed Abdelkader, Rafik Aguech . Moran random walk with reset and short memory. AIMS Mathematics, 2024, 9(8): 19888-19910. doi: 10.3934/math.2024971
    [5] He Song, Longmin Wang, Kainan Xiang, Qingpei Zang . The continuity of biased random walk's spectral radius on free product graphs. AIMS Mathematics, 2024, 9(7): 19529-19545. doi: 10.3934/math.2024952
    [6] H. M. Barakat, M. H. Dwes . Asymptotic behavior of ordered random variables in mixture of two Gaussian sequences with random index. AIMS Mathematics, 2022, 7(10): 19306-19324. doi: 10.3934/math.20221060
    [7] You Lv . Asymptotic behavior of survival probability for a branching random walk with a barrier. AIMS Mathematics, 2023, 8(2): 5049-5059. doi: 10.3934/math.2023253
    [8] Haoyang Ji, Wenxiu Ma . Phase transition for piecewise linear fibonacci bimodal map. AIMS Mathematics, 2023, 8(4): 8403-8430. doi: 10.3934/math.2023424
    [9] Yongfeng Wu . Limit theorems for negatively superadditive-dependent random variables with infinite or finite means. AIMS Mathematics, 2023, 8(11): 25311-25324. doi: 10.3934/math.20231291
    [10] Breix Michael Agua, Salim Bouzebda . Single index regression for locally stationary functional time series. AIMS Mathematics, 2024, 9(12): 36202-36258. doi: 10.3934/math.20241719
  • In this paper, we investigate an extended version of the elephant random walk model. Unlike the traditional approach where step sizes remain constant, our model introduces a novel feature: step sizes are generated as a sequence of positive independent and identically distributed random variables, and the step of the walker at time n+1 depends only on the steps of the walker between times 1,...,mn, where (mn)n1 is a sequence of positive integers growing to infinity as n goes to infinity. Our main results deal with the validity of the central limit theorem for this new variation of the standard ERW model introduced by Schütz and Trimper in 2004.



    The elephant random walk (ERW) represents a unique form of one-dimensional random walk along integers, notable for its retention of complete memory regarding its entire past trajectory. Schütz and Trimper [13] introduced the ERW as a means to investigate the memory effects inherent in a non-Markovian random walk. This model is inspired by the notion that elephants possess an exceptional memory, embodying the popular belief that elephants never forget their paths. It is well known that the asymptotic properties of the ERW depend on its memory parameter p[0,1]. More precisely, the ERW exhibits three regimes: the diffusive regime for 0p<3/4, the critical regime for p=3/4, and the superdiffusive regime of 3/4<p1. Many interesting limit theorems are already known for the ERW. In particular, n1Sn0 a.s. for any p]0,1[ and n1/2SnN(0,(34p)1) in distribution in the diffusive regime. In the critical regime, we have (nlogn)1/2SnN(0,1) in distribution and n12pSnL almost surely and in L4 in the superdiffusive regime. The reader can refer to Baur and Bertoin [2], Bercu [3,4], Bercu and Laulin [5], Kubota and Takei [10], Coletti, Gava, and Schütz [6], Laulin [11] and the references therein and to Ma et al. [12] and Dedecker et al. [7] for contributions to the rate of convergence in the central limit theorem of the ERW in the diffusive and critical regimes. Since the seminal paper by Schütz and Trimper [13], many variants of the ERW have been introduced in this literature. Aguech et al. [1] extended the results of Bercu [3] and Kubota and Takei [10] in the case where the memory of the elephant is increasing (see Gut and Stadtmuller [8]): At any step n, the elephant remembers only the steps at times 1,,mn, where mn is a non-decreasing integer sequence such that mn<n and limnmn/n=θ[0,1] and limnmn=+. In this paper, we combine the results in [7] and [1], and we assume that the size of the step of the elephant is random and with a non-decreasing memory mn. More precisely, suppose we have an ascending sequence of integers denoted as mn, where each mn is less than or equal to n. Let (Un)n1 be a sequence of uniformly distributed random variables on the set {1,,mn}, and let (Vn)n1 denote a sequence of random variables with values either -1 or 1 such that P(Vn=1)=p for some fixed p]0,1[. In [7], the authors assume that the size of the step of the elephant is a sequence (Zn)n1 of independent identically distributed random variables with finite variance σ2=V(Z1), and they assume that the sequences of random variables (Un)n1, (Vn)n1, and (Zn)n1 are independent. Denote by X1 the first step of the elephant and assume that X1 is a random variable such that P(X1=1)=1P(X1=1)=r for some fixed r[0,1]. For any integer n0, we denote by Sn the position of the elephant at time n. Following [7], we have S0=0, and for any integer n1,

    Sn=ni=1XiZi,

    where, for all n1,

    Xn+1=VnXUn={XUn, with probability p,XUn, with probability 1p,

    and recall that Un is uniformly distributed on {1,,mn} for any n1. Without loss of generality, one can assume that E(Zn)=1 for any n1, and consequently, the following decomposition will be usefull: For any n1,

    Sn=Tn+Hn,

    with

    Tn=nk=1XiandHn=nk=1Xi(Zi1).

    Denote a1=1, and for any n2,

    an=Γ(n)Γ(2p)Γ(n+2p1) and vn=ni=1a2i,

    where Γ(s)=0ts1etdt,s>0, is the Gamma function. For mn=n, Dedecker et al. [7] proved the following result:

    Theorem 1.1. [7, Theorems 2.1 and 3.1] Assume that p(0,1] and E[Z2]<+.

    (1) If p(0,3/4], then

    anSn(2r1)vn+na2nσ2dnN(0,1).

    (2) If p(3/4,1], then

    Snn2p1a.snL,

    where L is a non-denerate and non-Gaussian random variable.

    Remark 1.1. An important observation regarding the model investigated by Dedecker et al. [7] is its strong connection with the urn model. Specifically, the model operates as follows: Initially, an urn contains one white ball and one black ball. At each step n, a ball is randomly drawn from the urn. Subsequently, the drawn ball is replaced, with the addition of YnZn balls of the same color and (1Yn)Zn balls of the opposite color, where Yn has a Bernoulli distribution with parameter p and is independent of Zn. The addition matrix for this urn is random and is given by

    Dn=(YnZn(1Yn)Zn(1Yn)ZnYnZn).

    Some of the results of Dédecker et al. [7] and some extra results, can be obtained using this connection from [9], where, if we denote by Wn and Bn, respectively, the number of white and black balls at time n, the position of the elephant at step n can be given, due to this connection, by Sn=WnBn. This connection with the Pólya urn model is not possible if we consider gradually increasing memory random walks: Practically, it is not possible at step n to draw a ball only from the first mn added balls.

    It is important to note that Theorem 1.1 holds in the particular case of mn=n. Our aim in this work is to extend Theorem 1.1 to the case where mnn such that mn/nθ for some θ[0,1]. Additionally, for p>3/4 (superdiffusive case), we are going to give the asymptotic distribution for the fluctuation of the elephant random walk around the random variable τL, where L is defined in Theorem 1.1 and τ in Theorem 2.1.

    First case: Diffusive regime (0<p<3/4)

    Our first result concerns the case where p(0,3/4) and we give the asymptotic distribution of the elephant random walk.

    Theorem 2.1. Let θ[0,1] such that limnmn=+, and mn/nθ as n goes to infinity, and suppose that p(0,3/4). Assume that (Zn)n1 is iid positive with mean 1 and variance σ2, and denote

    τ=θ+(1θ)(2p1)andσ21=τ234p+θ(1θ).

    Then,

    mnnSndnN(0,σ21+θσ2).

    Remark 2.1. If the step size is deterministic, then σ=0, and we obtain the result in [1]. If θ=1, we obtain the result in [7].

    Proof of Theorem 2.1. Let

    ˉSn:=mnnSn=mnnHn+mnnTn=:An+Bn.

    By [1], we know that for 0<p<3/4, we have

    Bn=mnnTndnN(0,σ21), (2.1)

    with

    σ21=θ(1θ)+τ234pandτ2=θ+(1θ)(2p1).

    For all tR, denote φn(t)=E[eitˉSn] and consider the σ-algebra F=σ(Xi,Ui,Vi,iN). So, for any t in R, we have

    φn(t)=E[exp(it{An+Bn})]=E[E[exp(itAn)|F]exp(itBn)].

    Since the random variables (Zi1)i are iid centered with finite variance σ2 and that X2i=1 a.s., we get

    1nni=1Xi(Zi1)dnN(0,σ2). (2.2)

    In fact, if we denote

    Rn=1nni=1Xi(Zi1)andΦn(t)=E[exp(itRn)],

    then,

    Φn(t)=E[E[exp(itRn)F]]=E[nk=1E[exp(itnXk(Zk1))|F]]=E[nk=1E[exp(itnXk(Z11))|F]]=E[nk=1E[(1+itnXk(Z11)t22n(Z11)2+o(1n))|F]]=E[nk=1(1t2σ22n+o(1n))]=(1t2σ22n+o(1n))nexp(t2σ22),

    where in the last three equations, the o term depends on the moment of order two of Z1, which is finite by assumption.

    Using (2.2), we obtain

    An=mnnHnndnN(0,θσ2).

    As a first conclusion, we have

    limnE[exp(itAn)|F]=exp(θσ2t22)=φ(θσt).

    where φ is the characteristic function of the standard normal law. On the other hand, for any n1 and any tR,

    φn(t)=E[(E[exp(itAn)|F]φ(θσt))exp(itBn)]+φ(θσt)E[exp(itBn)].

    Using (2.1) and noting that

    limn+|E[(E[exp(itAn)|F]φ(θσt))exp(itBn)]|limnE[|E[exp(itAn)|F]φ(θσt)|]=0,

    we derive [1]

    limn+φn(t)=φ(θσt)φ(σ1t)=exp((θσ2+σ21)t22).

    The proof of Theorem 2.1 is complete.

    Second case: Critical regime (p=3/4)

    Assume that p=3/4 and consider the decompositon

    Sn=Smn+Dmn,n,whereDmn,n=nk=mn+1XkZk.

    Theorem 2.2. Let θ[0,1] such that mn/nθ as n goes to infinity, and suppose that p=3/4. If (Zn)n1 are iid positive with mean 1 and finite variance σ2, then

    mnSnnlnmndnN(0,(1+θ)24).

    Remark 2.2. Notice that in this case, the step size does not influence the asymptotic behavior of the normalized position of the elephant. This is due to the fact that the normalized remainder term Dmn,n converges to 0 in probability.

    Proof of Theorem 2.2. From [7], we know that

    SmnmnlnmndnN(0,1).

    Moreover, given Fmn:=σ(Xi;imn), for any kmn+1, we have

    P[Xk=1|Fmn]=(12+14mnTm) and P[Xk=1|Fmn]=(1214mnTm).

    Note also that, given Fmn, the random variables (XkZk)mn+1kn are i.i.d. In the other part, we have the following decomposition:

    ˜Sn:=mnSnnlnmn=An+Bn,

    where

    An:=mnnSmnmnlnmnandBn:=mnDmn,nnlnmn.

    For any tR and any n1, we denote ϕn(t)=E[exp(it˜Sn)] and Gn:=σ(Z1,,Zn). Then, we have

    ϕn(t)=E[E[exp(itAn)exp(itBn)|FmnGmn]]=E[exp(itAn)E[exp(itBn)|FmnGmn]]=E[exp(itAn)(E[exp(itmnnlnmnXmn+1Z1)|Fmn])(nmn)].

    Using the conditional distribution of Xmn+1 given Fmn, we obtain

    E[exp(itmnnlnmnXmn+1Z1)|Fmn]=(12+14mnTmn)E[exp(itmnnlnmnZ1)]+(1214mnTmn)E[exp(itmnnlnmnZ1)]=(12+14mnTmn)[1+itmnnlnmnμ1t2mnn2lnmnμ2+o(m3/2nn3lnm3/2n)]+(1214mnTmn)[1itmnnlnmnμ1t2mnn2lnmnμ2+o(m3/2nn3lnm3/2n)]=1+itμ12nTmnmnlnmnt2mnn2lnmnμ2+o(m3/2nn3lnm3/2n)+Tmnmn×o(m3/2nn3lnm3/2n),

    and recall that m1nTmna.s.n+0 (see [3]) and |m1nTmn|1 a.s. Then, for n sufficiently large

    (E[exp(itmnnlnmnXmn+1Z1)|Fmn])(nmn)=(1+itμ12nTmnmnlnmnt2mnn2lnmnμ2+(1+Tmnmn)×o(m3/2nn3lnm3/2n))(nmn)=exp[(nmn)ln(1+itμ12nTmnmnlnmnt2mnn2lnmnμ2+o(m3/2nn3))]=exp[itμ1(nmn)2nTmnmnlnmnt2mn(nmn)n2lnmnμ2+o(m3/2nn2)]exp[itμ1(nmn)2nTmnmnlnmn].

    Then, the characteristic function ϕn(t) has an asymptotic expression

    ϕn(t)=E[exp(itAn)exp(itμ1(nmn)2nTmnmnlnmn)]=E[exp(itmnn(Tmnmnlnmn+Hmnmnlnmn))exp(itμ1(nmn)2nTmnmnlnmn)]=E[exp(itmn+n2nTmnmnlnmn+itmnnHmnmnlnmn)]=E[exp(itmn+n2nTmnmnlnmn)E[exp(itmnnHmnmnlnmn)|F]].

    Since

    limnE[exp(itmnnHmnmnlnmn)|F]=1,

    we deduce that

    limnϕn(t)=limnE[exp(itmn+n2nTmnmnlnmn)].

    Finally, by [3], we conclude

    limnE[exp(itmn+n2nTmnmnlnmn)]=exp(t22(θ+1)24).

    The proof of Theorem 2.2 is complete.

    Third case: Superdiffusive regime (p>3/4)

    In this section, we consider the case, where p(3/4,1), called the superdifusive regime. First, we give the almost sure convergence of Sn.

    Theorem 2.3. Let θ[0,1] such that mn/nθ as n goes to infinity, and suppose that p(3/4,1). If (Zn)n1 are iid positive with mean 1 and finite variance σ2, then

    m2(1p)nnSna.sn(θ+(2p1)(1θ))L,

    where L is a non-Gaussian random variable.

    Proof of Theorem 2.3. Recall that

    Sn=Smn+Dmn,n,whereDmn,n=nk=mn+1XkZk,

    and consequently,

    m2(1p)nnSn=mnnSmnm2p1n+m2(1p)nnDmn,n.

    From [7], we know that

    mnnSmnm2p1na.snθL.

    On the other hand, according to Fmn, the sequence of random variables (XiZi)mn+1in is i.i.d., and for all imn+1,

    E[XiZi|Fmn]=E[Xi|Fmn]E[Zi]=(2p1)Tmnmn.

    Applying the strong law of large numbers condionally to Fm, for n large, we have

    Dmn,nnmn=1nmnnk=mn+1XkZk(2p1)Tmnmna.s.

    Then, almost surely, for n large

    m2(1p)nnDmn,n=nmnnm2(1p)n1nmnDmn,nnmnnm2(1p)nTmnmn(1θ)(2p1)L.

    The following result shows that the fluctuations of the elephant random walk are still gaussian around the random variable L as given in Theorem 2.3.

    Theorem 2.4. Let θ[0,1] such that mn/nθ as n goes to infinity, and suppose that p(3/4,1). If (Zn)n1 are iid positive with mean 1 and finite variance σ2, then

    m4p3n(Snm2(1p)nnτL)dnN(0,λ2+θσ2),

    where

    λ2=τ24p3τ(θ2p1θ)1p+θ(1θ)andτ=θ+(1θ)(2p1).

    Remark 2.3. This result generalizes the result obtained in [7], it precises in addition the asymptotic distribution of the fluctuation, and coincides with the result of Aguech et al. [1], in the case where σ=0.

    Proof of Theorem 2.4. We start again with the following decomposition:

    Sn=nk=1Xk+nk=1Xk(Zk1)=Tn+Hn.

    First, using [1], if m4p3n|n1mnθ|0, then

    m4p3n(Snm2(1p)nnτL)dnN(0,λ2).

    Moreover, we have

    m4p3n(Snm2(1p)nnτL)=m4p3n(Tnm2(1p)nnτL)+mnnnk=1Xk(Zk1):=Δn+˜Hn.

    Let, for all n, Hn=σ(X1,,Xn,L). For all real t, let ψZ(t)=E[exp(it[Z1])] the characteristic function of Z1 and Ψn(t) is the characteristic function of m4p3n(Snm2(1p)nnτL). So, we have

    Ψn(t)=E[exp(itΔn)exp(it˜Hn)]=E[exp(itΔn)E[exp(it˜Hn)|Hn]]=E[exp(itΔn)nk=1ψZ(mntXk)].

    If ψZ|Hn denotes the characteristic function of Z1 conditionally to Hn, then

    ψZ|Hn(mnntXk)=E[exp(mnnitXk(Z11))|Hn]=E[(1+itmnnXk(Z11)t22mnn2(Z11)2+o(m3/2nn3))|Hn](1t22mnn2σ2)(ifnislarge).

    Consequently, if n is large, then

    Ψn(t)=E[exp(itΔn)(1t22mnn2σ2)n]=E[exp(itΔn)exp(nk=1ln[1t22mnn2σ2])]E[exp(itΔn)]exp(t22mnnσ2)exp(t22[λ2+θσ2]).

    The proof of Theorem 2.4 is complete.

    In this work, we place emphasis on the fact that the asymptotic normality for the elephant random walk with gradually increasing memory and random step size still holds in the three regimes of the model. Our results extend previous ones established by Dedecker et al. [7] and Aguech et al. [1]. Additionally, we observed that the connection with Polya urns is not feasible for gradually increasing memory. Finally, we argue that our approach can be used for deriving many others limit theorems for the elephant random walk with gradually increasing memory and random step size (law of the iterated logarithm, rate of convergence in the central limit theorem, invariance principle).

    What about the asymptotic normality of the ERW model, remembering only its last mn steps nmn,,n1? This very interesting question will serve as the basis for a new research project. Another problem that we think is very interesting, and which will be one of our future projects, is to consider the same question but with an ERW model with a random step size. A more difficult problem is to establish the rate of convergence in the central limit theorems for the ERW model with restricted memory. In particular, it will be very interesting to understand the way that the restricted memory will influence the rate of convergence in the central limit theorem.

    The authors declare they have not used Artificial Intelligence (AI) tools in the creation of this article.

    The author would like to extend his sincere appreciation to Deanship of Scientific Research at King Saud University: Researchers Supporting Program number (RSPD2024R987).

    The author declares no conflict of interest.



    [1] R. Aguech, M. El Elmachkouri, Gaussian fluctuations of the elephant random walk with gradually increasing memory, J. Phys. A Math. Theor., 57 (2024), 065203. https://doi.org/10.1088/1751-8121/ad1c0d doi: 10.1088/1751-8121/ad1c0d
    [2] E. Baur, J. Bertoin, Elephant random walks and their connection to Pólya-type urns, Phys. Rev. E, 94 (2016), 052134. https://doi.org/10.1103/PhysRevE.94.052134 doi: 10.1103/PhysRevE.94.052134
    [3] B. Bercu, A martingale approach for the elephant random walk,, J. Phys. A Math. Theor., 51 (2018), 015201. https://doi.org/10.1088/1751-8121/aa95a6
    [4] B. Bercu, On the elephant random walk with stops playing hide and seek with the Mittag-Leffler distribution, J. Stat. Phys., 189 (2022), 12. https://doi.org/10.1007/s10955-022-02980-w doi: 10.1007/s10955-022-02980-w
    [5] B. Bercu, L. Laulin, On the multi-dimensional elephant random walk, J. Stat. Phys., 175 (2019), 1146–1163. https://doi.org/10.1007/s10955-019-02282-8 doi: 10.1007/s10955-019-02282-8
    [6] C. F. Coletti, R. Gava, G. M. Schutz., Central limit theorem and related results for the elephant random walk, J. Math. Phys., 58 (2017), 053303. https://doi.org/10.1063/1.4983566 doi: 10.1063/1.4983566
    [7] J. Dedecker, X. Q. Fan, H. J. Hu, F. Merlevede, Rates of convergence in the central limit theorem for the elephant random walk with random step sizes, J. Stat. Phys., 190 (2023), 154. https://doi.org/10.1007/s10955-023-03168-6 doi: 10.1007/s10955-023-03168-6
    [8] A. Gut, U. Stadtmuller, The elephant random walk with gradually increasing memory, Statist. Probab. Lett., 189 (2022), 109598. https://doi.org/10.1016/j.spl.2022.109598 doi: 10.1016/j.spl.2022.109598
    [9] S. Janson, Functional limit theorems for multitype branching processes and generalized Polya urns, Stochastic Process. Appl., 110 (2004), 177–245. https://doi.org/10.1016/j.spa.2003.12.002 doi: 10.1016/j.spa.2003.12.002
    [10] N. Kubota, M. Takei, Gaussian fluctuation for superdiffusive elephant random walks, J. Stat. Phys., 177 (2019), 1157–1171. https://doi.org/10.1007/s10955-019-02414-0 doi: 10.1007/s10955-019-02414-0
    [11] L. Laulin, Introducing smooth amnesia to the memory of the elephant random walk, Electron. Commun. Probab., 27 (2022), 1–12. https://doi.org/10.1214/22-ECP495 doi: 10.1214/22-ECP495
    [12] X. H. Ma, M. El Machkouri, X. Q. Fan, On Wasserstein-1 distance in the central limit theorem for elephant random walk, J. Math. Phys., 63 (2022), 013301. https://doi.org/10.1063/5.0050312 doi: 10.1063/5.0050312
    [13] G. Schutz, S. Trimper, Elephants can always remember: exact long-range memory effects in a non-Markovian random walk, Phys. Rev. E, 70 (2004), 045101. https://doi.org/10.1103/PhysRevE.70.045101 doi: 10.1103/PhysRevE.70.045101
  • This article has been cited by:

    1. Rafik Aguech, Elephant Random Walk with a Random Step Size and Gradually Increasing Memory and Delays, 2024, 13, 2075-1680, 629, 10.3390/axioms13090629
    2. Mohamed Abdelkader, Rafik Aguech, Moran random walk with reset and short memory, 2024, 9, 2473-6988, 19888, 10.3934/math.2024971
  • Reader Comments
  • © 2024 the Author(s), licensee AIMS Press. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Metrics

Article views(924) PDF downloads(46) Cited by(2)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog