Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of...

18
Research Article Moderate Deviations for Stochastic Fractional Heat Equation Driven by Fractional Noise Xichao Sun , 1 Ming Li , 2 and Wei Zhao 3 1 College of Science, Bengbu University, 1866 Caoshan Rd., Bengbu 233030, China 2 School of Information Science and Technology, East China Normal University, No. 500, Dong-Chuan Road, Shanghai 200241, China 3 American University of Sharjah, P.O. Box 26666, Sharjah, UAE Correspondence should be addressed to Xichao Sun; [email protected] Received 28 January 2018; Revised 6 April 2018; Accepted 2 May 2018; Published 10 July 2018 Academic Editor: Maricel Agop Copyright © 2018 Xichao Sun et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. We consider a class of stochastic fractional heat equations driven by fractional noises. A central limit theorem is given, and a moderate deviation principle is established. 1. Introduction Since the work of Freidlin and Wentzell [1], the large deviation principle (LDP) has been extensively developed for small noise systems and other types of models (such as interacting particle systems) (see [27]). Cardon-Weber [2] proved a LDP for a Burgers-type SPDE driven by white noise. Marquez-Carreras and Sarra [3] proved a LDP for a stochastic heat equation with spatially correlated noise, and Mellali and Mellouk [4] extended Marquez-Carreras and Sarras [3] to a fractional operator. Jiang et al. [5] proved a LDP for a fourth-order stochastic heat equation driven by fractional noise. Budhiraja et al. [6] studied large devia- tion properties of systems of weakly interacting particles. Budhiraja et al. [7] proved a large deviation for Brownian particle systems with killing. Similar to the large deviation, the moderate deviation problems also come from the theory of statistical inference. Using the moderate deviation principle (MDP), we can get the rate of convergence and an important method to construct asymptotic condence intervals, for example, Liming [8], Guillin and Liptser [9], Cattani and Ciancio [10], and other references therein. There are also many works about MDP about stochastic (partial) dierential equations; some surveys and literatures could be found in Budhiraja et al. [11], Wang and Zhang [12], Li et al. [13], Yang and Jiang [14], and the references therein. On the other hand, fractional equations have attracted many physicists and mathematicians due to various applications in risk management, image analysis, and statistical mechan- ics (see Droniou and Imbert [15], Bakhoum and Toma [16], Levy and Pinchas [17], Mardani et al. [18], Niculescu et al. [19], Paun [20], and Pinchas [21] for a survey of applica- tions). Stochastic partial dierential equations involving a fractional Laplacian operator have been studied by many authors; see Mueller [22], Wu [23], Liu et al. [24], Wu [25], and the references therein. Motived above, we investigated the moderate deviations about the stochastic fractional heat equation with fractional noise as follows: v ε t t , x = D δ,α v ε t , x + fv ε t , x + ε B H dt , dx , v ε o, x = 0, 1 where t 0, T , x D = 0, 1 , D δ,α is the fractional Laplacian operator which is dened in Appendix, and B H dt , dx denotes a fractional noise which is fractional in time and white in space with Hurst parameter H 1/2, 1 ; that is, B H is a mean zero Gaussian random eld on 0, T × D with covariance. Hindawi Complexity Volume 2018, Article ID 7402764, 17 pages https://doi.org/10.1155/2018/7402764

Transcript of Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of...

Page 1: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Research ArticleModerate Deviations for Stochastic Fractional Heat EquationDriven by Fractional Noise

Xichao Sun ,1 Ming Li ,2 and Wei Zhao3

1College of Science, Bengbu University, 1866 Caoshan Rd., Bengbu 233030, China2School of Information Science and Technology, East China Normal University, No. 500, Dong-Chuan Road, Shanghai 200241, China3American University of Sharjah, P.O. Box 26666, Sharjah, UAE

Correspondence should be addressed to Xichao Sun; [email protected]

Received 28 January 2018; Revised 6 April 2018; Accepted 2 May 2018; Published 10 July 2018

Academic Editor: Maricel Agop

Copyright © 2018 Xichao Sun et al. This is an open access article distributed under the Creative Commons Attribution License,which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

We consider a class of stochastic fractional heat equations driven by fractional noises. A central limit theorem is given, and amoderate deviation principle is established.

1. Introduction

Since the work of Freidlin and Wentzell [1], the largedeviation principle (LDP) has been extensively developedfor small noise systems and other types of models (suchas interacting particle systems) (see [2–7]). Cardon-Weber[2] proved a LDP for a Burgers-type SPDE driven by whitenoise. Marquez-Carreras and Sarra [3] proved a LDP for astochastic heat equation with spatially correlated noise, andMellali and Mellouk [4] extended Marquez-Carreras andSarra’s [3] to a fractional operator. Jiang et al. [5] proveda LDP for a fourth-order stochastic heat equation drivenby fractional noise. Budhiraja et al. [6] studied large devia-tion properties of systems of weakly interacting particles.Budhiraja et al. [7] proved a large deviation for Brownianparticle systems with killing.

Similar to the large deviation, the moderate deviationproblems also come from the theory of statistical inference.Using the moderate deviation principle (MDP), we canget the rate of convergence and an important method toconstruct asymptotic confidence intervals, for example,Liming [8], Guillin and Liptser [9], Cattani and Ciancio[10], and other references therein. There are also manyworks about MDP about stochastic (partial) differentialequations; some surveys and literatures could be foundin Budhiraja et al. [11], Wang and Zhang [12], Li et al.[13], Yang and Jiang [14], and the references therein. On

the other hand, fractional equations have attracted manyphysicists and mathematicians due to various applicationsin risk management, image analysis, and statistical mechan-ics (see Droniou and Imbert [15], Bakhoum and Toma [16],Levy and Pinchas [17], Mardani et al. [18], Niculescu et al.[19], Paun [20], and Pinchas [21] for a survey of applica-tions). Stochastic partial differential equations involving afractional Laplacian operator have been studied by manyauthors; see Mueller [22], Wu [23], Liu et al. [24], Wu[25], and the references therein.

Motived above, we investigated the moderate deviationsabout the stochastic fractional heat equation with fractionalnoise as follows:

∂vε

∂tt, x =Dδ,αv

ε t, x + f vε t, x + εBH dt, dx ,

vε o, x = 0,1

where t ∈ 0, T , x ∈D = 0, 1 , Dδ,α is the fractional Laplacianoperator which is defined in Appendix, and BH dt, dxdenotes a fractional noise which is fractional in time andwhite in space with Hurst parameter H ∈ 1/2, 1 ; thatis, BH is a mean zero Gaussian random field on 0, T ×Dwith covariance.

HindawiComplexityVolume 2018, Article ID 7402764, 17 pageshttps://doi.org/10.1155/2018/7402764

Page 2: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Cov BH t, x BH s, y =12

t2H + s2H − t − s 2H x∧y

2

Assume that the coefficients satisfy the following.

Assumption 1. Function f is Lipschitz; that is, there exist anm > 0 satisfying

f y − f x ≤m y − x , ∀x, y ∈D 3

Under the conditions of Assumption 1, (1) possesses aunique solution in the sense of Walsh [26] as follows:

vε t, x =t

0 DGδα t − s, x − z f vε s, z dzds

+ εt

0 DGδα t − s, x − z BH ds, dz

4

As the parameter ε→ 0, the solution vε t, x of(1) will tend to v0 t, x which is the solution to the fol-lowing equation:

∂v0

∂tt, x =Dδ,αv

0 t, x + f v0 t, x ,

v0 0, x = 05

This paper mainly devotes to investigate the deviations ofvε from the deterministic solution v0, as ε→ 0, that is, theasymptotic behavior of the trajectories.

Vε t, x ≔1εa ε

vε − v0 t, x ,  t, x ∈ 0, T ×D, 6

where a ε is the same deviation scale that strongly influ-ences the asymptotic behavior of Vε.

If a ε = 1/ ε, we are in the domain of large deviationestimate, which can be proved similarly to Jiang et al. [5].

The case a ε ≡ 1 provides the central limit theorem. Asε↓0, we will prove that vε − v0 / ε converges to a randomfield in this paper.

To fill the gap between scale a ε = 1 and scale a ε =1/ ε, we mainly devote to the moderate deviation whenthe scale satisfies the following:

a ε → +∞,  εa ε → 0, ε→ 0 7

This paper is organized as follows. In Section 2, thedefinition of the fractional noise BH ds, dz is given. InSection 3, the main result is given and proved. In Appendix,some results about the Green kernel are given.

2. Fractional Noise

Let H ∈ 1/2, 1 , and BH 0, t × A t,A ∈ 0,T ×B ℝ is acentered Gaussian family of random variables with thecovariance satisfying

E BH t, A BH s, B = A ∩ B RH t, s , 8

with s, t ∈ 0, T , A, B ∈B ℝ and covariance kernel RH

t, s = 1/2 t2H + s2H − t − s 2H , where ∣A∣ denotes theLebesgue measure of the set A ∈B ℝ and B ℝ denotesthe class of Borel sets in ℝ.

We denote φ as the set of step functions on 0, T ×ℝ. LetH be the Hilbert space defined as the closure of φ withrespect to the scalar product.

1 0,t ×A, 1 0,s ×BH= E BH t, A BH s, B 9

According to Nualart and Ouknine [27], the mapping1 0,t ×A → BH t, A can be extended to an isometry between

H and the Gaussian space H1 BH associated with BH anddenoted by

ϕ↦BH ϕ ≔t

0 Aϕ s, x BH ds, dx 10

Define the linear operator K∗H φ↦L2 0, T by

K∗H ϕ = KH T , s ϕ s, y +

T

sϕ r, y − ϕ s, y

∂KH

∂rr, s dr,

11

where KH is defined by

with CH = 2αΓ 3/2 − α /Γ α + 1/2 Γ 2 − 2α 1/2, andone can get

∂KH

∂tt, s = CH

12−H t − s H− 3/2 t

s

H− 1/213

Moreover, KH satisfies the following:

s∧t

0KH s, r KH t, r dr =

12

t2H + s2H − t − s 2H 14

KH t, s ≔CHs

1/2 −Ht

su − s H− 3/2 1 − s

u

1/2 −Hdu + CH t − s H− 1/2 , 0 < s ≤ t,

0, otherwise,12

2 Complexity

Page 3: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Then, since

K∗H1 0,t ×A s, x = KH t, s 1 0,t ×A s, x , 15

one can get

K∗H ϕ , K∗

H ψ L2 0,T ×D = ϕ, ψ ℋ, 16

where ϕ and ψ in φ are any step functions. So theoperator K∗

H gives an isometry between the Hilbert spaceℋ and L2 0, T × 0, 1 . Hence, W t, A , t ∈ 0, T ,A ∈ℬ 0, 1 defined by

W t, A = BH K∗H

−1 1 0,t ×A 17

is a space-time white noise, and BH has the followingform:

BH t, x =t

0

x

0KH t, s W ds, dy 18

Therefore, the mild formulation of (4) has the followingform:

vε t, x =t

0 DGδα t − s, x − z f vε s, z dzds

+ εt

0 DK∗

HGδα t − s, x − z W ds, dz

19

That is, the last term of (4) is equal to

t

0 DGδα t − s, x − z BH ds, dz

=t

0 DK∗

HGδα t − s, x − z W ds, dz

20

The following embedding proposition is given by Nualartand Ouknine [27].

Lemma 1. Set H > 1/2 , then, we have

L1/H 0, T ×ℝ ⊂ℋ 21

3. Main Results and Their Proof

3.1. Main Results. For any function ϕ defined on 0, T ×D,let

where μ ∈ 0, α + 1 H − 1/α , θ ∈ 0, min 1, α + 1 H − 1 ,and γ ∈ 0, 1 . Let

ϕ γ = ϕ T ,∞ + ϕ T ,γ 23

Let Cγ 0, T ×D be the functions ϕ 0, T ×D whichsatisfy ϕ γ <∞, endowed with the ϕ γ − norm.

Define

E = e x, t =t

0

x

0e s, z dsdz, e ∈ L2 0, T ×D , 24

which is a Cameron-Martin space endowed with the norm

e ℰ =t

0

x

0e2 s, z dzds

1/225

Suppose e ∈E. Now, let ve be the solution of the followingdeterministic equation:

∂ve

∂tt, x =Dδ,αv

e t, x + f ve t, x + F e t, x ,

ve 0, x = 0,26

where

F e x, t ≔∂2

∂x∂t

t

0

x

0KH t, s e z, s dzds 27

satisfies the following:

t

0 DGδα t − s, x − z F e z, e dzds

=t

0 DK∗

HGδα t − s, x − z e z, s dzds

28

The above (26) shows a unique mild solution.

Ze x, t =t

0 DGδα t − s, x − z f s, z, Ze z, s dzds

+t

0 DK∗

HGδα t − s, x − z e z, s dzds

29

Similar to Jiang et al. [5], one can get the following:

ϕ t,∞ = sup ϕ s, z : s, z ∈ 0, t ×D ,

ϕ t,γ = supϕ s1, z1 − ϕ s2, z2s1 − s2

μ + z1 − z2θ

γ s1, z1 ≠ s2, z2 ∈ 0, t ×D ,22

3Complexity

Page 4: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Theorem 1. Let H ∈ 1/2, 1 . Under Assumption1, the lawof the solution to (1) satisfies a deviation principle onCγ D × 0, T with the good rate function:

I φ = inf12

e 2ℰ ; Ze = φ , 30

with the convention inf∅ =∞, where 0 < γ < 1.More precisely, for any Borel measurable subset B of

Cγ 0, T ×D ,

−infφ∈Bo I φ ≤ lim infε→0

ε log P Ze ∈ B

≤ lim supε→0

ε log P Ze ∈ B ≤ −infφ∈BI φ ,31

where Bo and B denote the interior and the closure of B,respectively.

We furthermore suppose that the coefficients satisfy thefollowing.

Assumption 2. f is differentiable, and the derivative f ′ of f isLipschitz. That is to say, there exist positive constant m andm′ which satisfy the following:

f ′ y − f ′ x ≤m′ y − x , ∀x, y ∈D 32

Together with the Lipschitz of f , we conclude that

f ′ x ≤m 33

Now, we give the following central limit theorem.

Theorem 2. Let f and its derivative f ′ satisfy Assumptions 1and 2. Then, for p ≥ 1, ve − v0 / ε converges in Lp to a ran-dom field U on Cγ 0, T ×D with 0 < γ < 1, determined by

∂U t, x∂t

=Dδ,αU t, x + f ′ v0 t, x U t, x + BH dt, dx ,

U 0, x = 0,34

for all, t, x ∈ 0, T ×D.

Let the function Ue be the solution to the followingpartial differential equation:

∂Ue t, x∂t

=Dδ,αUe t, x + f ′ v0 t, x Ue t, x + F e t, x ,

Ue 0, x = 035

Under Assumptions 1 and 2, by Theorem 1, one canget U/a ε which satisfies large deviation principles onCγ 0, T ×D with the speed e2 ε and the good ratefunction satisfies the following:

I φinf

12

e 2ℰ, Ue = φ ,

+∞, otherwise36

Now, the second result is given as follows:

Theorem 3. In moderate deviation principle, let H ∈ 1/2, 1 .Under the Assumptions1and2, then, the random field 1/εa ε vε − v0 satisfies a large deviation principle on

the space Cγ 0, T ×D with speed a2 ε and the goodrate function I φ defined by (36), where 0 < γ < 1.

3.2. Convergence of the Solution

Lemma 2. Let H ∈ 1/2, 1 . Under Assumption1, then,there exists a unique solution to (1). Moreover, for anyp ∈ 1,∞ , T > 0,

supε≥1

supt,x ∈ 0,T ×D

E vε t, x p <∞ 37

As ε→ 0, we get the convergence of vε as follows:

Proposition 1. Let H ∈ 1/2, 1 and p ≥ 2. By Assumption 1,there exists a constant c m, p,H, T which satisfies thefollowing:

E vε − v0t,∞

p≤ εp/2c m, p,H, T 38

Proof. Note that

vε t, x − v0 t, x

=t

0 DGδα t − s, x − z f vε s, z − f v0 s, z dzds

+ εt

0 DGδα t − s, x − z BH dsdz

39

One can get

vε t, x − v0 t, xT ,∞

p≤ 2p−1 sup

t,x ∈ 0,T ×D

t

0 DGδα t − s, x − z f vε s, z − f v0 s, z dzds

p

+ εp/22p−1 supt,x ∈ 0,T ×D

t

0 DGδα t − s, x − z BH dsdz

p 40

4 Complexity

Page 5: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Set

Nε1 t, x =

t

0 DGδα t − s, x − z f vε s, z − f v0 s, z dzds,

Nε2 t, x =

t

0 DGδα t − s, x − z BH dsdz

41Together with Hölder’s inequality, the Lipschitz condi-

tion (C) and (5) of Lemma A.1, for 1/α < q < α, we have

E Nε1 T ,∞

p ≤mp supt,x ∈ 0,T ×D

Gδα t − s, x − z qdzds

p/q

× ET

0vε − v0

T ,∞

pdt

≤ C p,m, T ET

0vε − v0

T ,∞

pdt,

42where 1/p + 1/q = 1.

For any 0 ≤ t ≤ T , x, z ∈D, p > 2, by (21) and (A.9),there exist θ ∈ 0, min 1, α + 1 H − 1 which satisfiesthe following:

E Nε2 t, x −Nε

2 t, x′p

= Et

0 DGδα t − s, x − z − Gδ

α t − s, x′ − z BH dsdzp

≤t

0 DK∗

H Gδα t − s, x − z − Gδ

α t − s, x′ − z2dsdz

p/2

≤t

0 DGδα t − s, x − z −Gδ

α t − s, x′ − z1/H

dsdzHp

≤m x − x′ pθ

43

Similarly,

By (21), (A.10), and (A.11), for μ ∈ 0, α + 1 H − 1 /α , onecan get

t

t DK∗

H Gδα t − s, x − z

2dsdz

p/2

≤ CH

t

t DGδα t − s, x − z

1/Hdsdz

Hp

≤ CH t − t′p α+1 H−1 /α

45

and

t

0 DK∗

H Gδα t − s, x − z − Gδ

α t′ − s, x − z2dsdz

p/2

≤ CH

t

t DK∗

H Gδα t − s, x − z −Gδ

α t′ − s, x − z1/H

dsdzHp

≤ CH t − t′pμ

46

Together with (43), (44), (45), and (46), one can get forany p ≥ 2 and t′, x′ , t, x ∈ 0, T ×D, there exists aconstant β > 0 such that

E Nε2 t, x −Nε

2 t, x′p≤ C p,m,H, T x − x′ θ + t − t′ μ pβ

,

47

where C p,m,H, T is independent of ε. For p > 2/β, byGarsis-Rodemich-Rumser’s Lemma, there exist a constantC and a random variable Mp,ε ω satisfying

Nε2 t, x −Nε

2 t, x′p

≤mp,ε ω x − x′ θ + t − t′ μ pβ−2

⋅ logc

x − x θ + t − t μ

248

E Nε2 t, x −Nε

2 t′, xp= E

t

0 DGδα t − s, x − z − Gδ

α t′ − s, x − z BH dsdzp

≤t

0 DK∗

H Gδα t − s, x − z −Gδ

α t′ − s, x − z2dsdz

p/2

≤ C p,m, Tt

t DK∗

H Gδα t − s, x − z

2dsdz

p/2

+ C p,m, Tt

0 DK∗

H Gδα t − s, x − z −Gδ

α t − s, x − z2dsdz

p/2

44

5Complexity

Page 6: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

and

supε∈ 0,1

E mp,ε <∞ 49

If we choose t′ = 0 in (48), we can get

E supt,x ∈ 0,T ×D

t

0 DGδα t − s, x − z BH dsdz

p

≤ c m, p,H, T supε∈ 0,1

E mp,ε <∞50

By Gronwall’s inequality and (40), (42), and (48), thereexists a constant C p,m,H, T satisfying

E vε − v0t,∞

p≤ εp/2c m, p,H, T 51

The proof is completed.

3.3. The Proof of the Main Results.We first prove Theorem 1.Our proof is based on the following proposition (see Dossand Priouret [28]).

Proposition 2. Suppose Xϵj , ϵ > 0 and j = 1, 2 are two

families of random variables and Ej, dj and j = 1, 2 aretwo Polish spaces. Suppose that

(1) There exists a map K I1 <∞ → E1 such thatI1 ≤ a → E1 is continuous for any a <∞.

(2) Xϵ1, ϵ > 0 satisfies with the rate functions I1 E1 →

0,∞ .

(3) For any R, δ, a > 0, there exist ρ > 0 and ϵ0 > 0satisfying I1 ≤ a and ϵ ≤ ϵ0 for all h ∈ E1, and

P d1 Xϵ1 ≤ ρ, d2 xϵ2, K e ≥ δ ≤ exp −

Rϵ2

52

Then, Xϵ2, ϵ > 0 satisfies a LDP with the rate functions

I h = inf I1 e : K e = h 53

To prove Theorem 1, one only needs to prove

(i) Under some topology, Z e : e e E ≤ a →C 0, T , Lp D is continuous for any a > 0.

(ii) In Freidin-Wentzell inequality, for any R > 0, η > 0,and e ∈E, there exist a δ > 0 satisfying

Theorem 4. When the level set e E ≤ a is endowed withthe topology of uniform convergence on 0, T ×D ,

Z e : e eE≤ a → C 0, T , Lp D 55

is the continuous map for any a ∈ 0, +∞ .

Proof. One only needs to prove that for fixed a > 0, e, h ∈∥e∥E ≤ a ,

supt∈ 0,T ℝ

Z e t, x − Z h t, x 2dx ≤ C e − h ℰ 56

Note that

Using (A.5) in Appendix with p = q = 2, then ρ = 1, onecan get

limϵ→0

sup ϵ2logP ϵW − e ∞ < δ, supt∈ 0,T

uϵ t, · − Z e t, · ∥2 ≥ ρ ≤ −R 54

Z e t, x − Z h t, x =t

0 ℝGδα t − s, x − z f s, z, Z e s, z − f s, z, Z h s, z dzds

+t

0 ℝK∗

HGδα t − s, x − z e s, z dzds≔ I t, x + II t, x

57

I t, x 22 ≤ C

t

0t − s − 1/α + 1/αρ f s, z, Z e s, z − f s, z, Z h s, z 2ds ≤ C sup

0≤s≤tZ e s, ⋅ − Z h s, ⋅ 2 58

6 Complexity

Page 7: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Now, we deal with II t, x , together with

Recalling K∗H is defined by (11), one can get

K∗Hϕ s, x ≤ max

w∈ s,Tϕ w, x KH T , s ,

M t =ℝ

t

0 ℝK∗

HGδα t − s, x − z

2dzdsdx

≤ℝ

t

0 ℝmaxw∈ s,t

Gδα t − s, x − z

2K2

H t, s dzdxds

≤ Ct

0t − s − 1/α t − s 2H−1ds <∞

60

Using Gronwall’s inequality, we can get

supt∈ 0,T ℝ

Z e t, x − Z h t, x 2dx ≤ C e − h ℰ 61

The proof of the theorem is completed.

We now prove the Freidin-Wentizell inequality asfollows:

Suppose W t, x , t, x ∈ 0, T ×ℝ is a Brownian sheet.For e ∈E and ϵ > 0, we define

W t, x =W t, x −e x, tϵ 62

and

dpdp

= exp1ϵ

T

0 ℝe t, x W dt, dx −

12ϵ2

T

0 ℝe2 t, x dxdt ≔ ZT

63

Using Girsanov’s theorem, the process W is a Browniansheet under P. Suppose vε t, x is a solution of (1) underp. Then,

vϵ t, x =ℝGδα t, x − z v0 z dz

+t

0 ℝGδα t − s, x − z f t, x, vε s, z dzds

+t

0 ℝK∗

HGδα t − s, x − z e s, z dzds

+ ϵt

0 ℝK∗

HGδα t − s, x − z W ds, dz

64

Now, one can prove (36). Note that, under p, then,

So under p, by Gronwall’s Lemma, one can get

II t, x 22 =

t

0 ℝK∗

HGδα t − s, x − z e s, z − h s, z dzds

2dx

≤ Cℝ

t

0 ℝK∗

HGδα t − s, x − z

2dzdsdx × e − h 2

E ≔ CM t × e − h 2ℰ

59

uϵ t, · − Z e t, · 2 =t

0 ℝGδα t − s, · − z f t, x, uε s, z − f t, x, Z e s, z dzds

2

+ ϵt

0 ℝK∗

HGδα t − s, · − y W ds, dz

2

≤t

0 ℝuε s, · − Z e s, · 2dzds + ϵ

t

0 ℝK∗

HGδα t − s, · − z W ds, dz

2

65

supt∈ 0,T

uε s, · − Z h s, · 2 ≤ C supt∈ 0,T

ϵt

0 ℝK∗

HGδα t − s, x − z W ds, dz

266

7Complexity

Page 8: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Now, one can change (34) the proof to the followingtheorem.

Theorem 5. Suppose e ∈E and e ∈E ≤ a. For each R > 0,η > 0, and e ∈E, there exists a constant δ > 0 satisfying

In the following, we give a key Lemma to prove Theorem1, which is similar to Candon-Weber [2], and the proof isomitted.

Lemma 3. Suppose F 0, T ×D 2 →ℝ, CF > 0 and β0 > 0satisfying

T

0 DF t, x, v, z − F s, y, v, z 2dudz ≤ CF t − s μ + x − y θ

β0,

68

for any s, z and t, x ∈ 0, T ×D. Suppose N Ω ×0, T × 0, 1 →ℝ which is an almost surely continuous,F t-adapted process satisfying sup N t, x : t, x ∈ 0, T ×D ≤ ρ, a, s, and for t, x ∈ 0, T ×D, suppose

F t, x =T

0 DF t, x, v, z N v, z W dvdz 69

Then, for any 0 < β < β0/2 , there exist a positive constantC, Ç(β, β0), and C β, β0 such that for allM ≥ ρC1/2

F Ç(β, β0)C β, β0 ,

P F t,β ≥M ≤ C exp −M2

ρ2CzCβ, β0 C2 β, β070

Proof of Theorem 5. Suppose

F t, x, v, z = ϵK∗HG

δα t − v, x − z 71

Then, there exists β0 > 0 satisfying

If R ≤ ρ2/2CF C2 V and ρ ≤ K v C v CF , by Lemma2, we can get

P supt∈ 0,T

ϵt

0 ℝK∗

HGδα t − s, x − z W ds, dz

2≥ ρ

≤ 2 T2π2 + 1 exp −Rϵ2

73

The proof of Theorem 3 is completed.

In the following, we first give Garsis-Rodemich-Rumser’sLemma in Bally et al. [29].

Lemma 4. Assume p > 1 and Uε t, x : t, x ∈ 0, T ×Dare a family of real-valued stochastic processes. Suppose thefollowing is true.

Assumption 3. One has

limε→0

E Uε t, x p = 0 74

as t, x ∈ 0, T ×D.

limϵ→0

sup ϵ2 log P supt∈ 0,T

ϵt

0 ℝK∗

HGδα t − s, x − z W ds, dz

2≥ ρ, ϵW ∞ < δ ≤ −R 67

t

0 DK∗

H Gδα t − v, x − z −Gδ

α s − v, y − z2dvdz

≤ Ct

0 ℝK∗

H Gδα t − v, x − z − Gδ

α s − v, x − z2dvdz +

t

0 ℝK∗

H Gδα s − v, x − z − Gδ

α s − v, y − z2dvds

≤ Ct

o ℝGδα t − v, x − z −Gδ

α s − v, x − z1/H

dvds2H

+t

0 ℝGδα s − v, x − z −Gδ

α s − v, y − z1/H

dvdz2H

≤ C t − s μ + x − z θ β0

72

8 Complexity

Page 9: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Assumption 4. For any t, x , s, z ∈ 0, T ×D, there existsν > 0 satisfying

E U ε t, x −U ε s, z p ≤ C t − s + x − z 2 γ 75

Then, for any r ∈ 1, p , γ ∈ 0, β/p ,

limε→0

E Uεγ

r= 0 76

Now we can prove Theorem 2.

Proof of Theorem 2. Set Uε = vε − v0 / ε. We will prove that

limε→0

E U ε −U γ

r= 0 77

To this end, we need to prove that Assumptions 3 and 4are satisfied for Wε =U ε −U . Note that

Uε t, x −U t, x

=t

0 DGδα t − s, x − z

⋅f vε s, z − f v0 s, z

ε− f ′ v0 s, z U s, z dsdz

≕ IIIε t, x + IVε t, x78

where

IIIε t, x

=t

0 DGδα t − s, x − z

⋅f vε s, z − f v0 s, z

ε− f ′ v0 s, z U ε s, z dsdz

79

and

IVε t, x =t

0 DGδα t − s, x − z f ′ v0 s, z Uε s, z −U s, z dsdz

80

Using Taylor’s formula, there exists a ξε t, x such that

f vε t, x − f v0 t, x = f ′ v0 t, x + ξε t, x vε t, x − v0 t, x× vε t, x − v0 t, x

81

Note that f ′ is Lipschitz continuous and ξε t, x ∈ 0, 1 ;one can get

f ′ v0 t, x + ξε t, x vε t, x − v0 t, x × vε t, x − v0 t, x

≤m′ξε t, x vε t, x − v0 t, x ≤m′ vε t, x − v0 t, x

82

So

IIIε t, x ≤m′t

0 DGδα t − s, x − z vε t, x − v0 t, x Uε s, z dsdz

= εm′t

0 DGδα t − s, x − z Uε s, z 2dsdz

83

Using Hölder’s inequality, for p < 2 and 1/α < q < α, onecan get

E IIIε t∞

p ≤ εp/2m′p supt,x ∈ 0,T ×D

t

0 DGδα t − s, x − z

qdsdz

p/q

×t

0E Uε s

∞2pds,

84

where 1/p + 1/q = 1. Together with (5) of Lemma A.1 andProposition 1, there exists a constant C p,m,m′, T onlydepending on p,m,m′, T satisfying

E IIIε t∞

p ≤ εp/2C p,M,M , T 85

Since f ′ ≤M, for p > 2, together with Hölder’sinequality and (5) of Lemma A.1, we can get

E IVε t∞

p ≤Mp supt,x ∈ 0,T ×D

t

0 DGδα t − s, x − z

qdsdz

p/q

× Et

0Uε −U s

∞pds,

86

where 1/p + 1/q = 1. Together with (52), (56), and (57),we can get

E U ε −U t∞

p ≤ C p,m,m , T εp/2 + Et

0U ε −U s

∞pds

87

Using Gronwall’s inequality, one can get

E Uε −U t∞

p ≤ εp/2C p,m,m , T → 0, ε→ 0, 88

which implies Assumption 3 in Lemma 4.

9Complexity

Page 10: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Now we prove Assumption 4 in Lemma 4. We willprove IIIε and IVε satisfy Assumption 4 in Lemma 4.Using Hölder’s inequality and (A.9), one can get

where θ′ ∈ 0, min 1, α + 1 /q − 1 and 1/p + 1/q = 1.Similarly,

where μ ∈ 0, α + 1/q − 1 /α . Together with (89) and (90),we can get

E IIIε t, x − IIIε s, z p ≤ C t − s + x − z 2 β 91

Similarly, one can get

E IVε t, x − IVε s, z p ≤ C t − s + x − z 2 β 92

Together with (91) and (92), we can get

E Uε t, x −U t, x − Uε s, z −U s, z p

≤ C t − s + x − z 2 β 93

Then, for any p > 2 and q ∈ 1, α such that γ ∈ 0, β/pand r ∈ 1, p , using Lemma 4,

limε→0

E V εγ

r= 0 94

The proof of Theorem 2 is completed.

Proof of Theorem 3. By Theorem 1, U/a ε obeys largedeviation principles on Cγ 0, T ×D , with the rate functionI given by (30) and the speed function h2 ε . Using Demboand Zeitouni ([30] Theorem 4.2.13), to prove the largedeviation principles of Uε/a ε is e2 ε -exponentially equiva-lent to U/a ε , that is,

lim supε→0

e−2 ε log PU ε −U μ′,θ′

a ε> δ = −∞ 95

holds for any δ > 0.Since

U ε −U γ ≤ C T , µ′, θ′ Uε −U T ,γ 96

To prove (67), we only need to prove

lim supε→0

e−2 ε log PU ε −U t, μ′, θ′

a ε> δ = −∞ 97

Note the decomposition

Uε t, x −U t, x = IIIε t, x + IVε t, x 98

E IIIε t, x − IIIε t, xp≤ E εm

t

0 DGδα t − s, x − z − Gδ

α t − s, x − z Uε s, z 2dsdzp

≤ εp/2M pt

0 DGδα t − s, x − z −Gδ

α t − s, x ‐zqdsdz

p/q×

t

0 DE Yε s, z s

∞2pdsdz

1/p

≤ εp/2C p,m,m , T x − xpθ,

89

E IVε t, x − IVε t , xp≤ E εm

t

0 DGδα t − s, x − z −Gδ

α t ‐s, x‐z U ε s, z 2dsdzp

≤ εp/2C p,m, Tt

0 DGδα t − s, x − z −Gδ

α t ‐s, x‐z U ε s, z 2dsdzp

≤ εp/2C p,m, Tt

t DGδα t − s, x − z − Gδ

α t ‐s, x‐z U ε s, z 2dsdzp

+ εp/2C p,m, Tt

0 DGδα t − s, x − z Uε s, z

2

dsdz p ≤ C p,m,m , T t − tpμ,

90

10 Complexity

Page 11: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

For any q ∈ 1, α , x, x′ ∈ 0, 1 , 1/p + 1/q = 1, and0 ≤ s ≤ t ≤ T , by Hölder’s inequality, (32), and (A.9), wecan get

where θ′ ∈ 0, min 1, α + 1 /q − 1 . Similarly,

where μ′ ∈ 0, α + 1 /q − 1/α . Together with (99) and (100),we can get

E IVε t, x − IVε s, z p ≤ C t − s µ′ + x − z θ′ 101

Note that U ε −U s,∞ ≤ C s, µ′, θ′ Uε −U s,γ, we have

IVε t, x t,γ ≤ C C s, μ′, θ′ U ε −U s,γpds 102

Thus, for t ∈ 0, T , we have

Uε −U t,γp≤ C p,m, T IIIε t,γ +

t

0Uε −U s,γ

pds

103

Applying Gronwall’s Lemma to f t = Uε −U t,γp,

we have

U ε −U T ,γp≤ C p, T ,m IIIε T ,γ

p104

By (64) to prove (67), we only need to prove that for anyδ > 0,

lim supε→0

e−2 ε log PIε T,γ

a ε≥ δ = −∞ 105

Note that

IIIε t, x =t

0 DGδα t − s, x − z Aε s, z dsdz, 106

where

Aε s, z =f vε s, z − f v0 s, z

ε− f ′ v0 s, z U ε s, z

107

By the same method in the proof of (59), we have

IIIε T ,γ ≤ C T, µ′, θ′ AεT ,∞ 108

IVε t, x − IVε t, x′ ≤m′t

0 DGδα t − s, x − z −Gδ

α t − s, x′ − z Uε s, z −U s, z dsdz

≤m′t

0 DGδα t − s, x − z −Gδ

α t − s, x′ − zqdsdz

1/q×

t

0 DUε s, z −U s, z pdsdz

1/p

≤M′ x − x′ θ ×t

0U ε −U u

∞p

1/p

99

IVε t, x − IVε t′, x ≤m′t

0 DGδα t − s, x − z − Gδ

α t ‐s, x‐z U ε s, z −U s, z dsdz

≤m′t

0 DGδα t − s, x − z −Gδ

α t − s, x − zqdsdz

1/q×

t

0 DUε s, z −U s, z pdsdz

1/p

+m′t

t DGδα t − s, x − z

q

dsdz1/q

×t

t DU ε s, z −U s, z pdsdz

1/p

≤m′ t − t′ μ′ ×t

0Uε −U u

∞p

1/p

100

11Complexity

Page 12: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Similar to the proof of (53), we have

AεT ,∞ ≤m′

vε − v0T ,∞

2

ε109

Together with (40) and (A.11), for any t ∈ 0, T , we canget that

vε − v0t,∞

≤ supt,x ∈ 0 T ×D

l

0 DGδα l − s, x − z m vε s, z − v0 s, z dsdz

supt,x ∈ 0,T ×D

εl

0 DGδα t − s, x − z BH dsdz

≤ C m, Tt

0vε − v0

s,∞ds + IIVεt,∞,

110

where

IIVε t, x = ε1

0 DGδα t − s, x − z BH dsdz

= ε1

0 DK∗

HGδα t − s, x − z W dsdz

111

By Gronwall’s Lemma, one can get

vε − v0T ,∞ ≤ C M, T IIVε

T,∞ ≤ C M, T IIVε γ 112

Applying Lemma 4 with

F t, x, s, z = K∗HG t − s, x − z 1s≤t , β0 =

12, CF = C,

ρ = εM 1 + v0T ,∞ + θ ,N t, x = ε1 vε T ,∞< v0 T,∞+θ ,

113

for any fixed θ > 0, one can get that for any β,G ≥ εK 1 +v0∣T ,∞ + θ C1/2

F Ç β, β0 C β, β0 ,

P IIVε∣β,β ≥M, vε T ,∞ < v0 T,∞ + θ

≤ L exp −G2

εM2CC2 β, β0 1 + v0∣T ,∞ + θ 2

114

We can get

lim supε→0

e−2 ε log P vε∣T ,∞ ≥ v0T ,∞ + θ

≤ lim supε→0

e−2 ε log P vε − v0∣T ,∞ ≥ θ = −∞

115

Together with (73), (78), (82), (83), and (85), we have

This completes the proof.

Appendix

Green Function

The nonlocal factional differential operator Dδ,α isdefined by

ℱ Dδ,αϕ ξ = − ξ α exp −iδπ

2sgn ξ ℱ ϕ ξ , A 1

where α is called the index of stability and δ improperlyreferred to as the skewness which satisfy δ ≤min α − α 2,2 + α 2 − α and δ = 0 when δ ∈ 2ℕ + 1.

The operatorDδ,α is a closed, densely defined operator on

L2 ℝ , and it is the infinitesimal generator of a semigroupwhich is in general not symmetric and not a contraction. Thisoperator is a generalization of various well-known operators,such as the Laplacian operator (when α = 2), the inverse ofthe generalized Riesz-Feller potential (when α > 2), and the

lim supε→0

e−2 ε log PIII T,μ ,θ

a ε> δ

≤ lim supε→0

e−2 ε log P III T,μ ,θ2>

εa ε δ

C β, T, m

≤ lim supε→0

e−2 ε log P III T, μ , θ2>

εa ε δ

C β, T, m, vε T,∞ < v0 T,∞ + θ + P vε T ,∞ ≥ v0

T ,∞ + θ

≤ lim supε→0

−δεa ε C β, T ,m m2CC2 β, β0 1 + v0∣T ,∞ + θ 2 ∨ lim sup

ε→0e−2 ε log P vε∣T ,∞ ≥ v0

T ,∞ + θ = −∞

116

12 Complexity

Page 13: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Riemann-Liouville differential operator (when δ = 2 + α 2or δ = α − α ). It is self-adjoint only when δ = 0, and in thiscase, it coincides with the fractional power of the Laplacian.We refer the readers to Debbi and Dozzi [31] for more detailsabout this operator.

According to Komatsu [32], Dδ,α can be represented for1 < α < 2, by

Dδ,αφ x

=1

0

φ x + y − φ x − yφ′ xy 1+α κδ−1 −∞,0 y + κδ+1 −0,+∞ y dy,

A 2

and for 0 < α < 1, by

Dδ,αφ x =1

0

φ x + y − φ x

y 1+α κδ−1 −∞,0 y + κδ+1 −0,+∞ y dy,

A 3

where κδ− are κδ+ two nonnegative constants satisfying κδ− +κδ+ > 0, φ is a smooth function for which the integral exists,and φ′ is its derivative. This representation identifies it asthe infinitesimal generator for a nonsymmetric α-stableLévy process.

Suppose Gδα t, x is the fundamental solution to the

following equation:

∂v∂t

t, x =Dδ,α t, x ,

v 0, x = δ0 x , t > 0, x ∈ℝ,A 4

where δ0 · is the Dirac distribution. Using Fourier trans-form, one can get Gδ

α t, x which is given by

Gδα t, x =

12π D

exp −izx − t z α exp −iδπ

2sgn z dz

A 5

Let us list some known facts on Gδα t, x which will be used

later on (see, e.g., Debbi and Dozzi [31]).

Lemma A.1. Suppose α ∈ 0,∞ / ℕ , one can get thefollowing:

(1) Gδα t, · is not symmetric with respect to x.

(2) For any x ∈ℝ and s, t ∈ 0,∞ ,

∂n

∂xnGδα t, x = s − n+t /α Gδ

α s−1t, s− 1/α x A 6

or equivalently

∂n

∂xnGδα t, x = t − n+t /α Gδ

α 1, t − 1/α x A 7

(3) For any s, t ∈ 0,∞ , Gδα s, ⋅ ∗Gδ

α t, ⋅ =Gδα s + t, ⋅

(4) For n ≥ 1, there exist constants C and Cn > 0satisfying

Gδα 1, x ≤ C

11 + x 1+α , 

∂n

∂xnGδα 1, x ≤ Cn

x α+n−1

1 + x α+n 2

A 8

(5) T0 D Gδ

α t, x λdtdx <∞ if and only if 1/α < λ < α.

Lemma A.2. For θ ∈ 0, min 1, α + 1 H − 1 and μ ∈0, α + 1 H − 1 /α , we have

t

0 ℝGδα t − v, y −w − Gδ

α t − v, y −w1/H

dwdvH

≤m x − y θ,

A 9

t

0 ℝGδα t − v, x −w −Gδ

α s − v, x −w1/H

dwdvH

≤m t − s μ,

A 10

t

s ℝGδα t − v, x −w

1/Hdwdv

H

≤m t − s α+1 H−1 /α,

A 11

where m > 0 is a constant.

Proof. For any x, y ∈ℝ and t ∈ 0, T ,

t

0 ℝGδα t − v, x −w − Gδ

α t − v, y −w1/H

dwdvH

= Gδα t − ∗, x − ⋅ −Gδ

α t − ∗, y − ⋅L1/H 0,T ×ℝ

= Gδα t − ∗, x − ⋅ −Gδ

α t − ∗, y − ⋅θ

⋅ Gδα t − ∗, x − ⋅ − Gδ

α t − ∗, y − ⋅1−θ

L1/H 0,T ×ℝ

≤ Gδα t − ∗, x − ⋅ −Gδ

α t − ∗, y − ⋅θ

⋅ Gδα t − ∗, x − ⋅

1−θ

L1/H 0,T ×ℝ

+ Gδα t − ∗, x − ⋅ −Gδ

α t − ∗, y − ⋅θ

⋅ Gδα t − ∗, y − ⋅

1−θ

L1/H 0,T ×ℝ≔ I1 + I2

A 12

13Complexity

Page 14: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

By the mean-value theorem, for θ ∈ 0, 1 , one canget that

and

∂∂x

Gδα t − u, η −w

θ/HGδα t − u, x −w

1−θ /Hdw

= t − u 1/α − 1+θ /αH

∂∂x

Gδα 1, η −w

θ/HGδα 1,w

1−θ /Hdw

≤ CH,θ t − u 1/α − 1+θ /αH

A 14

Therefore, if 1/α − 1 + θ /αH + 1 > 0, that is, θ <α + 1 H − 1, then,

I1 ≤ CT ,θ,H x −w θ A 15

Similarly, we can check

I2 ≤ CT ,θ,H x −w θ A 16

Hence, the inequality (A.9) holds. As for the inequality(A.10), for any x ∈ℝ and t, s ∈ 0, T ,

t

0 ℝGδα t − v, x −w − Gδ

α s − v, x −w1/H

dzdvH

= Gδα t − ∗, x − ⋅ −Gδ

α s − ∗, x − ⋅L1/H 0,T ×ℝ

= Gδα t − ∗, x − ⋅ − Gδ

α s − ∗, x − ⋅μ

⋅ Gδα t − ∗, x − ⋅ −Gδ

α s − ∗, x − ⋅1−μ

L1/H 0,T ×ℝ

≤ Gδα t − ∗, x − ⋅ −Gδ

α s − ∗, x − ⋅μ

⋅ Gδα s − ∗, x − ⋅

1−μ

L1/H 0,T ×ℝ

+ Gδα t − ∗, x − ⋅ −Gδ

α s − ∗, x − ⋅μ

⋅ Gδα s − ∗, x − ⋅

1−μ

L1/H 0,T ×ℝ≔ II1 + II2

A 17

By mean-value theorem, it holds that

Note that

Gδα t, y = t−1/αGδ

α 1, t− 1/α y A 19

Hence, one can attain

∂∂t

Gδα t, y = −

t− 1− 1/α Gδα 1, t− 1/α y

+ −1α

t− 1−2 1/α y∂∂z

Gδα 1,w w=t− 1/α y

A 20

So

I1 = x − y θ ∂∂x

Gδα t − ∗, η − ⋅

θ

⋅ Gδα t − ∗, x − ⋅

1−θ

L1/H 0,T ×ℝ

= x − y θ/H∂

0 ℝ

∂∂x

Gδα t − u, η −w

θ/HGδα t − v, x −w

1−θ /Hdwdu

HA 13

II1 = t − s μ ∂∂t

Gδα ρ − ∗, x − ⋅

μ

⋅ Gδα t − ∗, x − ⋅

1−μ∥L1/H 0,T ×ℝ

=T

0 ℝt − s ∣ μ/H

∂∂t

Gδα ρ − r, x −w

μ/HGδα t − r, x −w

1−μ /H

dwdu

H A 18

14 Complexity

Page 15: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Therefore, if − αμ + 1 /Hα + 1/α + 1 > 0 i.e. μ <α + 1 H − 1 /α, then, for μ ∈ 0, a + 1 H − 1 /α ,

II1 ≤ Cα,H,T,µ t − s µ A 22

Similarly,

II2 ≤ Cα,H,T,µ t − s µ A 23

So (A.10) holds. Now let us prove that (A.11) holds. Forany x ∈ℝ and s, t ∈ 0, T ,

t

s ℝGδα t − v, x −w

1/Hdwdu

H

≤t

st − u − 1/αH + 1/α

ℝGδα 1, z

1/Hdw dv

H

≤ CH t − s a+1 H−1 /α

A 24

The proof of the lemma is completed.

Lemma A.3. Suppose p ∈ 1,∞ , q ∈ 1, p , and ρ ∈ 1,∞satisfying

1ρ=1p−1q+ 1 ∈ 0, 1 A 25

Suppose Gδα = Gδ

α t, x − z is the Green kernel, Q = ∂/∂zGδα orQ = Gδ

α with t, x, z ∈ 0, T ×ℝ ×ℝ. We define I by

I u t, x =t

0 ℝQ t − s, x − z u s, z dsds, A 26

with u ∈ L1 0, T ; Lq . Then, I L1 0, T ; Lq →L∞ 0, T ; Lq is a bounded linear operator which satisfiesthe following:

(1) When Q = ∂/∂y Gδα, then, for all r ∈ 1, 1 + α/2 ,

I u t, ⋅ p

≤ Ct

0t − s − 2/α + 1/αr u x, ⋅ qds, ∀t ∈ 0, T

A 27

(2) When Q = Gδα, then, for all ρ ∈ 1, 1 + α ,

I u t, ⋅ p

≤ Ct

0t − s − 1/α + 1/αr u x, ⋅ qds, ∀t ∈ 0, T

A 28

Proof. Together withMinkowski’s inequality, Young inequal-ity and (4) of Lemma A.1. We can get

I u t, ⋅ p =t

0 ℝ

∂∂z

Gδα t − s, ⋅ − z v s, z dzds

p

≤t

0 ℝ

∂∂z

Gδα t − s, ⋅ − z u s, z dz

p

ds

≤ Ct

0t − s − 2/a

∂∂z

Gδα 1, t − s − 1/a ⋅ −z u s, z dz

p

ds

≤ Ct

0t − s − 2/a

∂∂z

Gδα 1, t − s − 1/a ⋅ ∗ u s, ⋅ ⋅

p

ds

≤ Ct

0t − s − 2/a ∂

∂yGδα 1, t − s − 1/a ⋅

p

⋅ u s, ⋅ qds

≤ Ct

0t − s − 2/a + 1/αρ u s, ⋅ qds,

A 29

∂∂t

Gδα ρ − r, x −w

μ/HGδα t − r, x −w

1−μ /Hdw

≤ Cα,Hℝt − r −1− 1/α μ/H − 1/α 1−u /H Gδ

α 1, ρ − u − 1/α x −wμ/H

⋅ Gδα 1, t − u − 1/α x −w

1−μ /Hdw

+ Cα,Hℝt − r −1− 2/α μ/H − 1/α 1−u /H x −w μ/H ∂

∂wGδα 1,w ∣w= ρ−r − 1/α x−w

μ/H⋅ Gδ

α 1, t − u − 1/α x −w1−μ /H

dw

≤ Cα,H t − u − αμ+1 /Hα + 1/α + Cα,H t − u − αμ+μ+1 /Hα + 1/α

A 21

15Complexity

Page 16: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

where one has used the result that, for ρ ∈ 1, 1 + α /2 ,

Gδα 1, t − s − 1/α

ρ=

∂∂z

Gδα 1, t − s − 1/α z

ρdz

1/p

≤ t − s 1/αρ

∂∂y

Gδα 1, y

ρdy

1/p

≤ C t − s 1/αρ

A 30

The proof of (2) is omitted since it is similar to case (1).The proof of this lemma is completed.

Data Availability

No data were used to support this study.

Conflicts of Interest

The authors declare that there is no conflict of interestregarding the publication of this article.

Acknowledgments

Xichao Sun is partially supported by the Natural ScienceFoundation of Anhui Province (1808085MA02) and the Nat-ural Science Foundation of Bengbu University (2017GJPY04,2016KYTD02, 2017jxtd2, and 2017ZR08). Ming Li is par-tially supported by National Natural Science Foundation ofChina under the project Grant nos. 61672238, 61272402,and 61070214.

References

[1] M. I. Freidlin and A. D. Wentzell, Random Perturbationsof Dynamical Systems, Springer-Verlag, New York, NY,USA, 1984.

[2] C. Cardon-Weber, “Large deviations for a Burgers’-typeSPDE,” Stochastic Processes and their Applications, vol. 84,no. 1, pp. 53–70, 1999.

[3] D. Marquez-Carreras and M. Sarra, “Large deviation principlefor a stochastic heat equation with spatially correlated noise,”Electronic Journal of Probability, vol. 8, no. 12, 2003.

[4] T. El Mellali andM. Mellouk, “Large deviations for a fractionalstochastic heat equation in spatial dimension ℝd driven by aspatially correlated noise,” Stochastics and Dynamics, vol. 16,no. 1, article 1650001, 2016.

[5] Y. M. Jiang, K. H. Shi, and Y. J. Wang, “Large deviationprinciple for the fourth-order stochastic heat equations withfractional noises,” Acta Mathematica Sinica, English Series,vol. 26, no. 1, pp. 89–106, 2010.

[6] A. Budhiraja, P. Dupuis, and M. Fischer, “Large deviationproperties of weakly interacting processes via weak conver-gence methods,” The Annals of Probability, vol. 40, no. 1,pp. 74–102, 2012.

[7] A. Budhiraja, W. T. Fan, and R. Wu, “Large deviations forBrownian particle systems with killing,” Journal of TheoreticalProbability, vol. 6, pp. 1–40, 2016.

[8] W. Liming, “Moderate deviations of dependent randomvariables related to CLT,” The Annals of Probability, vol. 23,no. 1, pp. 420–445, 1995.

[9] A. Guillin and R. Liptser, “Examples of moderate deviationprinciple for diffusion processes,” Discrete and ContinuousDynamical Systems - Series B, vol. 6, no. 4, pp. 803–828, 2006.

[10] C. Cattani and A. Ciancio, “On the fractal distribution ofprimes and prime-indexed primes by the binary image anal-ysis,” Physica A: Statistical Mechanics and its Applications,vol. 460, pp. 222–229, 2016.

[11] A. Budhiraja, P. Dupuis, and A. Ganguly, “Moderate deviationprinciples for stochastic differential equations with jumps,”The Annals of Probability, vol. 44, no. 3, pp. 1723–1775, 2016.

[12] R. Wang and T. Zhang, “Moderate deviations for stochasticreaction-diffusion equations with multiplicative noise,” Poten-tial Analysis, vol. 42, no. 1, pp. 99–113, 2015.

[13] Y. Li, R. Wang, and S. Zhang, “Moderate deviations for astochastic heat equation with spatially correlated noise,” ActaApplicandae Mathematicae, vol. 139, no. 1, pp. 59–80, 2015.

[14] J. Yang and Y. Jiang, “Moderate deviations for fourth-orderstochastic heat equations with fractional noises,” Stochasticsand Dynamics, vol. 16, no. 6, article 1650022, 2016.

[15] J. Droniou and C. Imbert, “Fractal first-order partial differen-tial equations,” Archive for Rational Mechanics and Analysis,vol. 182, no. 2, pp. 299–331, 2006.

[16] E. G. Bakhoum and C. Toma, “Alternating coordinate-momentum representation for quantum states based on boppoperators for modelling long-distance coherence aspects,”Mathematical Problems in Engineering, vol. 2015, Article ID818019, 7 pages, 2015.

[17] C. Levy and M. Pinchas, “Maximum likelihood estimation ofclock skew in IEEE 1588 with fractional Gaussian noise,”Mathematical Problems in Engineering, vol. 2015, Article ID174289, 24 pages, 2015.

[18] A. Mardani, M. R. Hooshmandasl, M. H. Heydari, andC. Cattani, “A meshless method for solving the time fractionaladvection–diffusion equation with variable coefficients,” Com-puters & Mathematics with Applications, vol. 75, no. 1,pp. 122–133, 2018.

[19] O. Niculescu, D. G. Dimitriu, V. P. Paun, P. D. Matasaru,D. Scurtu, and M. Agop, “Experimental and theoretical inves-tigations of a plasma fireball dynamics,” Physics of Plasmas,vol. 17, no. 4, article 042305, 2010.

[20] V. Paun, “Fractal surface analysis of Zircaloy-4 SEM micro-graphs using the time-series method,” Central EuropeanJournal of Physics, vol. 7, no. 2, 2009.

[21] M. Pinchas, “Symbol error rate for nonblind adaptiveequalizers applicable for the SIMO and FGn case,”Mathemat-ical Problems in Engineering, vol. 2014, Article ID 606843,11 pages, 2014.

[22] C. Mueller, “The heat equation with Lévy noise,” StochasticProcesses and their Applications, vol. 74, no. 1, pp. 67–82, 1998.

[23] J. Wu, “Fractal Burgers equation with stable Lévy noise,”in International Conference SPDE and Applications-VII,pp. 4–10, 2004.

[24] J. Liu, L. Yan, and Y. Cang, “On a jump-type stochastic frac-tional partial differential equation with fractional noises,”Nonlinear Analysis: Theory, Methods & Applications, vol. 75,no. 16, pp. 6060–6070, 2012.

[25] D. Wu, “On the solution process for a stochastic fractionalpartial differential equation driven by space–time white noise,”

16 Complexity

Page 17: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Statistics & Probability Letters, vol. 81, no. 8, pp. 1161–1172, 2011.

[26] J. B. Walsh, “An introduction to stochastic partial differentialequations,” in École d'Été de Probabilités de Saint FlourXIV – 1984, vol. 1180 of Lecture Notes in Mathematics,pp. 265–439, Springer, Berlin, Heidelberg, 1986.

[27] D. Nualart and Y. Ouknine, “Regularization of quasilinear heatequations by a fractional noise,” Stochastics and Dynamics,vol. 4, no. 2, pp. 201–221, 2004.

[28] H. Doss and P. Priouret, “Petites perturbations de systemesdynamiques avec reflexion,” in Séminaire de Probabilités XVII1981/82, vol. 986 of Lecture Notes in Mathematics, pp. 353–370, Springer, Berlin, Heidelberg, 1983.

[29] V. Bally, A. Millet, and M. Sanz-Sole, “Approximation andsupport theorem in Holder norm for parabolic stochasticpartial differential equations,” The Annals of Probability,vol. 23, no. 1, pp. 178–222, 1995.

[30] A. Dembo and O. Zeitouni, Large Deviations Techniques andApplications, Springer-Verlag, Second edition, 1998.

[31] L. Debbi and M. Dozzi, “On the solutions of nonlinearstochastic fractional partial differential equations in onespatial dimension,” Stochastic Processes and their Applications,vol. 115, no. 11, pp. 1764–1781, 2005.

[32] T. Komatsu, “On the martingale problem for generators ofstable processes with perturbations,” Osaka Journal of Mathe-matics, vol. 21, pp. 113–132, 1984.

17Complexity

Page 18: Research Articledownloads.hindawi.com/journals/complexity/2018/7402764.pdfget the rate of convergence and an important method to construct asymptotic confidence intervals, for example,

Hindawiwww.hindawi.com Volume 2018

MathematicsJournal of

Hindawiwww.hindawi.com Volume 2018

Mathematical Problems in Engineering

Applied MathematicsJournal of

Hindawiwww.hindawi.com Volume 2018

Probability and StatisticsHindawiwww.hindawi.com Volume 2018

Journal of

Hindawiwww.hindawi.com Volume 2018

Mathematical PhysicsAdvances in

Complex AnalysisJournal of

Hindawiwww.hindawi.com Volume 2018

OptimizationJournal of

Hindawiwww.hindawi.com Volume 2018

Hindawiwww.hindawi.com Volume 2018

Engineering Mathematics

International Journal of

Hindawiwww.hindawi.com Volume 2018

Operations ResearchAdvances in

Journal of

Hindawiwww.hindawi.com Volume 2018

Function SpacesAbstract and Applied AnalysisHindawiwww.hindawi.com Volume 2018

International Journal of Mathematics and Mathematical Sciences

Hindawiwww.hindawi.com Volume 2018

Hindawi Publishing Corporation http://www.hindawi.com Volume 2013Hindawiwww.hindawi.com

The Scientific World Journal

Volume 2018

Hindawiwww.hindawi.com Volume 2018Volume 2018

Numerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisNumerical AnalysisAdvances inAdvances in Discrete Dynamics in

Nature and SocietyHindawiwww.hindawi.com Volume 2018

Hindawiwww.hindawi.com

Di�erential EquationsInternational Journal of

Volume 2018

Hindawiwww.hindawi.com Volume 2018

Decision SciencesAdvances in

Hindawiwww.hindawi.com Volume 2018

AnalysisInternational Journal of

Hindawiwww.hindawi.com Volume 2018

Stochastic AnalysisInternational Journal of

Submit your manuscripts atwww.hindawi.com