Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])

We present here an application of the results on simulation of weakly self-similar stationary increment φ-sub-Gaussian processes, obtained by Kozachenko, Sottinen and Vasylyk in [1], to the process of fractional Brownian motion.

Gespeichert in:
Bibliographische Detailangaben
Datum:2006
Hauptverfasser: Kozachenko, Y., Vasylyk, O.
Format: Artikel
Sprache:Englisch
Veröffentlicht: Інститут математики НАН України 2006
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/4457
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1]) / Y. Kozachenko, O. Vasylyk // Theory of Stochastic Processes. — 2006. — Т. 12 (28), № 3-4. — С. 55–62. — Бібліогр.: 4 назв.— англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1860066062564851712
author Kozachenko, Y.
Vasylyk, O.
author_facet Kozachenko, Y.
Vasylyk, O.
citation_txt Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1]) / Y. Kozachenko, O. Vasylyk // Theory of Stochastic Processes. — 2006. — Т. 12 (28), № 3-4. — С. 55–62. — Бібліогр.: 4 назв.— англ.
collection DSpace DC
description We present here an application of the results on simulation of weakly self-similar stationary increment φ-sub-Gaussian processes, obtained by Kozachenko, Sottinen and Vasylyk in [1], to the process of fractional Brownian motion.
first_indexed 2025-12-07T17:07:42Z
format Article
fulltext Theory of Stochastic Processes Vol. 12 (28), no. 3–4, 2006, pp. 55–62 YURIY KOZACHENKO AND OLGA VASYLYK SIMULATION OF FRACTIONAL BROWNIAN MOTION WITH GIVEN RELIABILITY AND ACCURACY IN C([0, 1])1 We present here an application of the results on simulation of weakly self-similar stationary increment ϕ-sub-Gaussian processes, obtained by Kozachenko, Sottinen and Vasylyk in [1], to the process of fractional Brownian motion. 1. Introduction In this paper we consider simulation of fractional Brownian motion de- fined on the interval [0, 1] with given reliability and accuracy in space C([0, 1]). We apply results obtained in [1] for centred second order Subϕ(Ω)-pro- cesses defined on the interval [0, 1] with covariance function R(t, s) = 1 2 ( t2H + s2H − |t − s|2H ) . The parameter H takes values in the interval (0, 1). In order to construct a model of such process we used a series expansion approach based on a series representation proved by Dzaparidze and van Zanten [2] for the fractional Brownian motion B: (1) Bt = ∞∑ n=1 sin(xnt) xn Xn + ∞∑ n=1 1 − cos(ynt) yn Yn, t ∈ [0, 1]. Here the Xn’s and the Yn’s are independent zero mean Gaussian random variables with certain variances depending on H and n. The xn’s are the positive real zeros of the Bessel function J−H of the first kind and the yn’s are the positive real zeros of the Bessel function J1−H . The series in (1) converge in mean square as well as uniformly on [0, 1] with probability 1. Replacing the Xn’s and Yn’s by independent random variables from the space Subϕ(Ω) we got series representation for ϕ-subGaussian random pro- cesses with covariance function R. This representation was used for sim- ulation of such processes with given reliability and accuracy in C([0, 1]). Processes of fractional Brownian motion belong to the space Subϕ(Ω) with 1Invited lecture. 2000 Mathematics Subject Classification. Primary 60G18, 60G15, 68U20, 33C10. Key words and phrases. Fractional Brownian motion, ϕ-sub-Gaussian processes, series expansions, simulation. 55 56 YURIY KOZACHENKO AND OLGA VASYLYK ϕ(x) = x2/2. So, in this paper we present some examples of simulation of fractional Brownian motion with different values of parameter H. 2. Space Subϕ(Ω) We need the following facts about the space Subϕ(Ω) of ϕ-sub-Gaussian (or generalised sub-Gaussian) random variables. Definition 2.1 ([3]). A continuous even convex function u = {u(x), x ∈ R} is an Orlicz N-function if it is strictly increasing for x > 0, u(0) = 0, u(x) x → 0 as x → 0 and u(x) x → ∞ as x → ∞. Proposition 2.2 ([3]). The function u is an Orlicz N-function if and only if u(x) = |x|∫ 0 l(u) du, x ∈ R, where the density function l is nondecreasing, right continuous, l(u) > 0 as u > 0, l(0) = 0 and l(u) → ∞ as u → ∞. Definition 2.3. Let u be an Orlicz N-function. The even function u∗ = {u∗(x), x ∈ R} defined by the formula u∗(x) = sup y>0 ( xy − u(y) ) , x ≥ 0, is the Young-Fenchel transformation of the function u. Proposition 2.4 ([3]). The function u∗ is an Orlicz N-function and for x > 0 u∗(x) = xy0 − u(y0) if y0 = l−1(x). Here l−1 is the generalised inverse function of l, i.e. l−1(x) := sup{v ≥ 0 : l(v) ≤ x}. Definition 2.5. An Orlicz N-function ϕ satisfies assumption Q if ϕ is qua- dratic around the origin, i.e. there exist such constants x0 > 0 and C > 0 that ϕ(x) = Cx2 for |x| ≤ x0. Definition 2.6. A zero mean random variable ξ belongs to the space Subϕ(Ω), the space of ϕ-sub-Gaussian random variables, if there exists a positive and finite constant a such that the inequality E exp{λξ} ≤ exp{ϕ(aλ)} holds for all λ ∈ R. SIMULATION OF FRACTIONAL BROWNIAN MOTION 57 The space Subϕ(Ω) is a Banach space with respect to the norm τϕ(ξ) = inf { a ≥ 0 : E exp{λξ} ≤ exp { ϕ(aλ) } , λ ∈ R } . Definition 2.7. A stochastic process X = (Xt)t∈[0,1] is a Subϕ(Ω)-process if it is a bounded family of Subϕ(Ω) random variables: Xt ∈ Subϕ(Ω) for all t ∈ [0, 1] and sup t∈[0,1] τϕ(Xt) < ∞. The properties of random variables from the spaces Subϕ(Ω) were studied in the book [4]. Remark 2.8. When ϕ(x) = x2 2 the space Subϕ(Ω) is called the space of sub- Gaussian random variables and is denoted by Sub(Ω). Centred Gaussian random variable ξ belongs to the space Sub(Ω), and in this case τϕ(ξ) is just the standard deviation: (Eξ2)1/2. Also, if ξ is bounded, i.e. |ξ| ≤ c a.s. then ξ ∈ Sub(Ω) and τϕ(ξ) ≤ c. 3. Simulation of Subϕ(Ω)-processes Define a process Z = (Zt)t∈[0,1] by the expansion (2) Zt = ∞∑ n=1 cn sin(xnt) ξn + ∞∑ n=1 dn ( 1 − cos(ynt) ) ηn, where cn = πH √ 2c xH+1 n J1−H(xn) , n = 1, 2, . . . ,(3) dn = πH √ 2c yH+1 n J−H(yn) , n = 1, 2, . . . ,(4) c = Γ(2H + 1) sin(πH) π2H+1 ,(5) ξn, ηn, n = 1, 2, . . . , are independent identically distributed centred random variables from the space Subϕ(Ω) with Eξ2 n = Eη2 n = 1 and τϕ(ξn) = τϕ(ηn) =: aϕ, n = 1, 2, . . . ; xn is the nth positive real zero of the Bessel function J−H ; yn is the nth positive real zero of J1−H , Jν(x) = ∞∑ n=0 (−1)n(x/2)ν+2n Γ(n + 1)Γ(ν + n + 1) . 58 YURIY KOZACHENKO AND OLGA VASYLYK Here x > 0, ν �= −1,−2, . . . and Γ denotes the Euler Gamma function Γ(z) = ∫ ∞ 0 tz−1e−t dt. We shall assume that the function ϕ( √ · ) is convex. Since ϕ-sub-Gaussian random variables are square integrable we have the following direct consequence of the series representation (1) for fractional Brownian motion. Proposition 3.1. The series (2) converges in mean square and the covari- ance function of the process Z is R. Theorem 3.2 ([1]). The series (2) converges uniformly with probability one and the process Z is almost surely continuous on [0, 1]. Moreover, if Z is strongly self-similar with stationary increments then it is β-Hölder continuous with any index β < H. Consider the space C([0, 1]) equipped with the usual sup-norm. Definition 3.3. The model Z̃ approximates the process Z with given reli- ability 1 − ν, 0 < ν < 1, and accuracy δ > 0 in C([0, 1]) if P ( sup t∈[0,1] |Zt − Z̃t| > δ ) ≤ ν. Let c̃n and d̃n be the approximated values of the cn and dn, respectively. Let |c̃n − cn| ≤ γc n, |d̃n − dn| ≤ γd n, n = 1, . . . . The errors γc n and γd n are assumed to be known. Let x̃n and ỹn be approxi- mations of the corresponding zeros xn and yn with error bounds |x̃n − xn| ≤ γx n, |ỹn − yn| ≤ γy n. The error bounds γx n and γy n are also assumed to be known. Then, the model of the process Z we define as follows (6) Z̃t = N∑ n=1 ( c̃n sin(x̃nt) ξn + d̃n ( 1 − cos(ỹnt) ) ηn ) . The following theorem contains the main result of the paper [1]. SIMULATION OF FRACTIONAL BROWNIAN MOTION 59 Theorem 3.4 ([1]). Let b and α be such that 0 < b < α < H. Denote γ0 = √ γappr + γcut, γα = √ γappr α + γcut α , β = min { γ0, γα 2α } , where γcut = a2 ϕ ∞∑ n=N+1 ( c2 n + 4d2 n ) , γappr = a2 ϕ N∑ n=1 {( cnγx n + γc n )2 + ( dnγy n + 2γd n )2 } , γcut α = 22−2αa2 ϕ ∞∑ n=N+1 ( c2 nx2α n + d2 ny2α n ) , γappr α = 23−2αa2 ϕ N∑ n=1 { x2α n (γc n)2 + y2α n (γd n)2 + 23−2α ( (c̃n)2(γx n)2α ( (xn + x̃n)2α 22α + 1 ) + + (d̃n)2(γy n)2α ( (yn + ỹn) 2α 22α + 1 ))} . Let l be the density of ϕ. The model Z̃, defined by (6), approximates the separable process Z, defined by (2), with given reliability 1 − ν, 0 < ν < 1, and accuracy δ > 0 in C([0, 1]) if the following three inequalities are satis- fied: (7) γ0 < δ, (8) βγ0 γα < δ 2α(exp{ϕ(1)} − 1)α , (9) 2 exp { −ϕ∗ ( δ γ0 −1 )}( 1 2b(1− b α ) ( γαδ βγ0 ) b α l−1 ( δ γ0 −1 ) +1 )2 b ≤ ν. 60 YURIY KOZACHENKO AND OLGA VASYLYK 4. Simulation of fractional Brownian motion Let us assume that the constants cn and dn and the zeros xn and yn are correctly calculated. In case of sub-Gaussian random processes we have the following corollary from the Theorem 3.4. Corollary 4.1. Suppose that there is no approximation error, i.e. γc n = γd n = γx n = γy n = 0. If the process Z is sub-Gaussian then conditions of the Theorem 3.4 are satisfied if (10) N ≥ max ⎧⎨ ⎩ ( aϕ δ √ 5c 2H )1/H + 1; 22− 4 H 5 1 H π ⎫⎬ ⎭ and (11) 2μ exp ⎧⎨ ⎩−1 2 ⎛ ⎝ δNH aϕ √ 5c 2H −1 ⎞ ⎠ 2⎫⎬ ⎭N14 ≤ ν, where μ = π22 22 H −45− 8 H ( H c ) 6 H ( δ aϕ ) 12 H . Recall that in the sub-Gaussian case we have ϕ(x) = x2 2 and that centered Gaussian random variables belong to the space Sub(Ω). Parameters α and b have to be optimized, but here α = H 2 and b = H 4 . If in the series representation (2) the ξn and ηn, n = 1, 2, . . . , are inde- pendent identically distributed centered Gaussian random variables with Eξ2 n = Eη2 n = 1, n = 1, 2, . . . , then Z is a process of fractional Brownian motion. In this case aϕ = 1. Using Corollary 4.1 we construct a model Z̃ of the fractional Brownian motion Z, such that Z̃ approximates Z with given reliability 1 − ν = 0.99 and accuracy δ = 0.01 in C([0, 1]). In this paper we present examples of such models for three values of parameter H . • H1 = 3 4 . In this case we have: c = 0.0537337; μ = 97700.7. From condition (10) follows that N ≥ max {101.5, 0.270008} . And condition (11) gives us N ≥ 6832.5. So, we take N = 6833. As a result of simulation we have the model, which approximates fractional Brownian motion with para- meter H = 3 4 with given reliability 0.99 and accuracy 0.01 in C([0, 1]) (see Figure 1). SIMULATION OF FRACTIONAL BROWNIAN MOTION 61 Figure 1. Model of fractional Brownian motion with parameter H = 3 4 • H2 = 7 8 . In this case c = 0.0264278; μ = 50832.8. From conditions (10) and (11) follows: N ≥ max {57.8064, 0.336974} and N ≥ 1054.22. The model with N = 1055 for fractional Brownian motion with parameter H = 7 8 is presented on Figure 2. Figure 2. Model of fractional Brownian motion with parameter H = 7 8 • H3 = 8 9 . In this case we have: c = 0.0234107 and μ = 47369.5. From condition (10) follows that N ≥ max {54.8061, 0.344046} . From condition (11) we have N ≥ 863.771. For N = 864 we obtained the model presented on Figure 3. We can see that, as it was expected, N decreases when value of H in- creases, so the closer is H to 1 the smoother curve we get. All calculations and simulation were made using software Mathematica. Acknowledgments The authors were supported by the European Union project Tempus Tacis NP 22012-2001 and the NATO Grant PST.CLG.980408. 62 YURIY KOZACHENKO AND OLGA VASYLYK Figure 3. Model of fractional Brownian motion with parameter H = 8 9 References 1. Kozachenko, Yu., Sottinen, T., Vasylyk, O., Simulation of Weakly Self- Similar Stationary Increment Subϕ(Ω)-Processes: A Series Expansion Approach, Methodology and Computing in Applied Probability, 7, (2005), 379–400. 2. Dzhaparidze, K. O. and van Zanten, J. H., A series expansion of frac- tional Brownian motion, Probability Theory and Related Fields, 130, (2004), 39–55. 3. Krasnoselskii, M. A. and Rutitskii, Ya. B. , Convex Functions in the Orlicz spaces, “Fizmatiz”, Moscow, (1958). 4. Buldygin, V. V. and Kozachenko, Yu. V., Metric Characterization of Random Variables and Random Processes, American Mathematical Society, Providence, RI, (2000). Department of Probability Theory and Math. Statistics, Mechanics and Mathematics faculty, Taras Shevchenko Kyiv National University, Volodymyrska 64, Kyiv, Ukraine E-mail address: yvk@univ.kiev.ua E-mail address: vasylyk@univ.kiev.ua
id nasplib_isofts_kiev_ua-123456789-4457
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0321-3900
language English
last_indexed 2025-12-07T17:07:42Z
publishDate 2006
publisher Інститут математики НАН України
record_format dspace
spelling Kozachenko, Y.
Vasylyk, O.
2009-11-11T15:20:16Z
2009-11-11T15:20:16Z
2006
Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1]) / Y. Kozachenko, O. Vasylyk // Theory of Stochastic Processes. — 2006. — Т. 12 (28), № 3-4. — С. 55–62. — Бібліогр.: 4 назв.— англ.
0321-3900
https://nasplib.isofts.kiev.ua/handle/123456789/4457
We present here an application of the results on simulation of weakly self-similar stationary increment φ-sub-Gaussian processes, obtained by Kozachenko, Sottinen and Vasylyk in [1], to the process of fractional Brownian motion.
en
Інститут математики НАН України
Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
Article
published earlier
spellingShingle Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
Kozachenko, Y.
Vasylyk, O.
title Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
title_full Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
title_fullStr Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
title_full_unstemmed Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
title_short Simulation of fractional Brownian motion with given reliability and accuracy in C([0, 1])
title_sort simulation of fractional brownian motion with given reliability and accuracy in c([0, 1])
url https://nasplib.isofts.kiev.ua/handle/123456789/4457
work_keys_str_mv AT kozachenkoy simulationoffractionalbrownianmotionwithgivenreliabilityandaccuracyinc01
AT vasylyko simulationoffractionalbrownianmotionwithgivenreliabilityandaccuracyinc01