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

Application of neural networks to inverse elastic scattering problems with near-field measurements

  • This paper is concerned with the application of a machine learning approach to inverse elastic scattering problems via neural networks. In the forward problem, the displacements are approximated by linear combinations of the fundamental tensors of the Cauchy-Navier equations of elasticity, which are expressed in terms of sources placed inside the elastic solid. From the near-field measurement data, a two-layer neural network method consisting of a gated recurrent unit to gate recurrent unit has been used to reconstruct the shape of an unknown elastic body. Moreover, the convergence of the method is proved. Finally, the feasibility and effectiveness of the presented method are examined through numerical examples.

    Citation: Yao Sun, Lijuan He, Bo Chen. Application of neural networks to inverse elastic scattering problems with near-field measurements[J]. Electronic Research Archive, 2023, 31(11): 7000-7020. doi: 10.3934/era.2023355

    Related Papers:

    [1] Teh Raihana Nazirah Roslan, Sharmila Karim, Siti Zulaiha Ibrahim, Ali Fareed Jameel, Zainor Ridzuan Yahya . Stochastic pricing formulation for hybrid equity warrants. AIMS Mathematics, 2022, 7(1): 398-424. doi: 10.3934/math.2022027
    [2] Guiwen Lv, Ping Xu, Yanxue Zhang . Pricing of vulnerable options based on an uncertain CIR interest rate model. AIMS Mathematics, 2023, 8(5): 11113-11130. doi: 10.3934/math.2023563
    [3] Din Prathumwan, Thipsuda Khonwai, Narisara Phoochalong, Inthira Chaiya, Kamonchat Trachoo . An improved approximate method for solving two-dimensional time-fractional-order Black-Scholes model: a finite difference approach. AIMS Mathematics, 2024, 9(7): 17205-17233. doi: 10.3934/math.2024836
    [4] Min-Ku Lee, Jeong-Hoon Kim . Closed-form approximate solutions for stop-loss and Russian options with multiscale stochastic volatility. AIMS Mathematics, 2023, 8(10): 25164-25194. doi: 10.3934/math.20231284
    [5] Ashish Awasthi, Riyasudheen TK . An accurate solution for the generalized Black-Scholes equations governing option pricing. AIMS Mathematics, 2020, 5(3): 2226-2243. doi: 10.3934/math.2020147
    [6] Xiuxian Chen, Zhongyang Sun, Dan Zhu . Mean-variance investment and risk control strategies for a dynamic contagion process with diffusion. AIMS Mathematics, 2024, 9(11): 33062-33086. doi: 10.3934/math.20241580
    [7] Kazem Nouri, Milad Fahimi, Leila Torkzadeh, Dumitru Baleanu . Numerical method for pricing discretely monitored double barrier option by orthogonal projection method. AIMS Mathematics, 2021, 6(6): 5750-5761. doi: 10.3934/math.2021339
    [8] Chao Yue, Chuanhe Shen . Fractal barrier option pricing under sub-mixed fractional Brownian motion with jump processes. AIMS Mathematics, 2024, 9(11): 31010-31029. doi: 10.3934/math.20241496
    [9] Xun Lu, Wei Shi, Changhao Yang, Fan Yang . Exponential integral method for European option pricing. AIMS Mathematics, 2025, 10(6): 12900-12918. doi: 10.3934/math.2025580
    [10] Kanagaraj Muthuselvan, Baskar Sundaravadivoo, Kottakkaran Sooppy Nisar, Fahad Sameer Alshammari . New technique for solving the numerical computation of neutral fractional functional integro-differential equation based on the Legendre wavelet method. AIMS Mathematics, 2024, 9(6): 14288-14309. doi: 10.3934/math.2024694
  • This paper is concerned with the application of a machine learning approach to inverse elastic scattering problems via neural networks. In the forward problem, the displacements are approximated by linear combinations of the fundamental tensors of the Cauchy-Navier equations of elasticity, which are expressed in terms of sources placed inside the elastic solid. From the near-field measurement data, a two-layer neural network method consisting of a gated recurrent unit to gate recurrent unit has been used to reconstruct the shape of an unknown elastic body. Moreover, the convergence of the method is proved. Finally, the feasibility and effectiveness of the presented method are examined through numerical examples.



    Modeling derivative products in mathematical finance usually starts with a system of stochastic differential equation that correspond to state variables like stock, interest rate and volatility. For financial analysts, economic auditors and risk managers in financial organizations, securities analysis is very important, thus in last decades they have tried to design new models that help to change markets react and solve some problems in the market. Modeling challenges has created a common link between applied mathematicians and financial researchers as the production of knowledge in this area [4,15].

    As an important model in the original Black–Scholes option valuation theory the volatility and the interest rate are both assumed constant. A popular expansion of the Black–Scholes model has been proposed by Heston. This expansion allows also for correlation between the volatility and the asset price, so that one can capture skewness effects [16]. In addition to the volatility the values of many practically important securities are sensitive to interest rates, necessitating the construction of interest rate models [17].

    Consider an agent who at each time has a portfolio valued at X(t). This portfolio invests in a money market account paying a constant rate of interest r, and in a stock modeled by the geometric Brownian motion,

    dS(t)=αS(t)dt+δS(t)dW(t), (1.1)

    Suppose at each time t, the investor holds Δ(t) shares of stock. The position Δ(t) can be random but must be adapted to the filtration associated with the Brownian motion W(t),t0. The remainder of the portfolio value, X(t)Δ(t)S(t), is invested in the money market account [12].

    In addition to, we consider the following model which presents the system of three SDEs which they have composed from different assets called hybrid model. The asset price model in following can be viewed as an extension of the popular Heston stochastic volatility model where the interest rate is not constant but also follows a stochastic process, described here by the Hull–White model.

    dSτ=RτSτdτ+SτVτdW1τ,S(0)>0, (1.2)
    dVτ=k(ηVτ)dτ+δ1VτdW2τ,V(0)>0,
    dRτ=a(b(τ)Rτ)dτ+δ2RτdW3τ,R(0)>0,

    In this model for 0<τT with T>0 the given maturity time of an option. The stocks process Sτ,Vτ, and Rτ denote random variables that represent the asset price, its variance, and the interest rate, at time τrespectively. The variance process Vτ was proposed by Heston and interest rate process Rτ, by Cox, Ingersoll and Ross in Eq.(1.2). Parameters k,η,δ1,δ2, and a, are given positive real numbers and dW1τ,dW2τ,dW3τ are Brownian motions with given correlation factors ρ1,ρ2,ρ3[1,1]. Also b, is a given deterministic positive function of time which can be chosen such that to match the current term structure of interest rates [12,13].

    The introduction of Legendre wavelet method for the variational problems by Razzaghi and Yousefi in 2000 and 2001 [3,11,14], several works applying this method were born. The Legendre wavelets method (LWM) is transformed a boundary value problem (BVP) into a system of algebraic equations [14]. The unknown parameter of this system is the vector that its components are the decomposition coefficients of the BVP solution into Legendre wavelets basis. In this paper we apply the multi- Legendre wavelet method to solve B-S and HCIR partial differential equations.

    This paper is organized as follows: In section 2, PDE models of Black-Scholes equation and Heston-Cox-Ingersoll-Ross equation have derived from SDE models, Section 3 has described properties and formulation of Legendre wavelet and Multi-dimensional Legendre Wavelet, Section4, metodology of multi-dimensional LWM is applied on PDEs that use in section 2. In section 5, numerical results and figures are presented and section 6, Conclusion is used to review.

    In his section we discuss about the B-S PDE model and HCIR PDE model that they have derived from SDE models [1,2,19].

    Theorem 2.1. (Ito-Doeblin formula for Brownian motion). Let f(t,x) be a function for which the partial derivatives ft(t,x),fx(t,x) and fxx(t,x) are defined and continues and let W(t) be a Brownian motion. Then for every T0,

    f(t,W(t))=f(0,W(0))+T0ft(t,W(t))dt (2.1)
    +T0fx(t,W(t))dt+T0fxx(t,W(t))dt,

    Proof: see [6].

    Remark 2.1: In the Ito- Doeblin formula we can write in differential form [6],

    df(t,W(t))=ft(t,W(t))+ft(t,W(t))dW(t)+12ft(t,W(t))dW(t)dW(t),+ftx(t,W(t))dtdW(t)+12fxx(t,W(t))dtdt, (2.2)

    But

    dW(t)dW(t)=d(t),d(t)dW(t)=dW(t)d(t)=0,d(t)d(t)=0, (2.3)

    And the Ito- Doeblin formula in differential form simplifies to

    df(t,W(t))=ft(t,W(t))dt+fx(t,W(t))dW(t)+12fxxdt. (2.4)

    Definition 2.1: let W(t),t0 be a Brownian motion and let F(t),t0 be an associated filtration. An Ito process as the form following

    X(t)=X(0)+t0(u)dW(u)+t0θ(u)d(u), (2.5)

    Where X(0) is nonrandom and (u) and θ(u) are adapted stochastic process.

    Lemma 2.1: the quadratic variation of the Ito process Eq.(2.5) is

    [X,X](t)=t02(u)du. (2.6)

    Proof: see [6].

    Definition 2.2: let X(t),t0 be an Ito process as described in Definition 2.1 and let Γ(t),t0 be an adapted process. We define the integral with respect to an Ito process,

    t0Γ(u)dX(u)=t0Γ(u)(u)dW(u)+t0Γ(u)θ(u)du. (2.7)

    Theorem 2.2. (Ito-Doeblin formula for an Ito process). Let X(t),t0 be an Ito process as described in Definition 2.1 and let f(t,x) be a function for which the partial derivatives ft(t,x),fx(t,x) and fxx(t,x) are defined and continues. Then for every T0,

    f(T,X(T))=f(0,X(0))+0(T)ft(t,X(t))dt (2.8)
    +T0fx(t,X(t))dX(t)+12T0fxx(t,X(t))d[X,X](t)=
    f(0,X(0))+T0ft(t,X(t))dt+T0fx(t,X(t))(t)dW(t)dt
    +T0fx(t,X(t))θdt+12T0fxx(t,X(t))(t)2(t)

    We may rewrite Eq.(2.8),

    df(t,x(t))=ft(t,X(t))dt+fx(t,X(t))dX(t)+12fxx(t,X(t))dX(t)dX(t). (2.9)

    proof: See [6].

    A SDE has defined as te following

    dX(u)=β(u,X(u))du+γ(u,X(u))dW(u). (2.10)

    The functions β(u,X(u)) and γ(u,X(u)) called drift and diffusion respectively. Where t0 final goal in these problems is to find a stochastic procesX(T) which is governed by

    X(t)=x, (2.11)
    X(t)=X(t)+Ttβ(u,X(u))du+Ttγ(u,X(u))dW(u), (2.12)

    Under conditions of the functions β(u,x) and γ(u,x) there exists a unique process X(T),Tt Satisfying Eq.(2.11) and Eq.(2.12) and generally this is a SDE form of following

    dX(u)=(a(u)+b(u)X(u))du+(γ(u)+δ(u)X(u))dW(u), (2.13)

    Where a(u),b(u),δ(u) and γ(u) are non-random functions of time and we permit on the right- hand side of Eq.(2.10) is the randomness inherent in the solution X(u) and in the driving Brownian motion W(u).Then the stochastic differential equation for geometric Brownian motion is

    dS(u)=αS(u)du+γS(u)dW(u), (2.14)

    The rate of change dSS, can be decomposed in to two parts in Eq.(2.14): One is deterministic, the other one is random.

    The deterministic part can be modeled by dSS=μdt, where μ is a measure of the growth rate of the asset and random part modeled by a Brownian motion δdW(t) that δ is the variance of the return and is called the volatility. The overall asset price model is then given by

    dSS=μdt+γdW(t), (2.15)

    Eq.(2.15), is a SDE and it can be expressed as follows

    dS(t)=μ(S(t))dt+δ(S(t))dWs(t). (2.16)

    An important lemma for finding their solution is the following Ito lemma.

    Lemma 2.2: Suppose S(t) satisfies the stochastic differential Eq.(2.16), and u(S,t) is a smooth function. Then u(S(t),t) satisfies the following stochastic differential equation [8].

    du=(ut+μus+12δ2uss)dt+δusdWs. (2.17)

    By expanding Eq.(2.17) with applying different variables in Ito lemma to earn B-S PDE and HCIR PDE equations, as following Black-Scholes PDE equation [1].

    Option transaction contract is divided into different types which the most important types are call option and put option. call option contract means that buyer have to buy the certain asset with given price in certain time, as the same way the put option contract leads to sale of asset with given price and certain time by seller. Maturity validity includes American option and European option [5]. Where W(t), is a standard Brownian motion. We also assume that interest rate are constant so that one unit of currency invested in the cash account at time 0, will be worth

    B(t)=exp(rt),

    at time t. We will value of call option or put option at time t. By Ito Lemma we know that,

    du(S,t)=(μS(t)uS+ut+12δ2S2(t)2uS2)dt+δS(t)uSdW(t), (2.18)

    Let us now consider a self-financing trading strategy where at each time t, we hold units xt, of the cash account and yt, units of the stock. Then pt, value of this strategy satisfies,

    pt=xtBt+ytSt, (2.19)

    We will choose xt and yt in such a way that the strategy replicates the value of the option the self-financing assumption implies that

    dpt=xtdBt+ytdSt, (2.20)
    =rxtBtdt+yt(μStdt+δStdWt) (2.21)
    =(rxtBtdt+ytμSt)dt+ytδStdWt,

    Note that Eq.(2.20) is consistent with our earlier definition of financing. In particular gains or losses on the portfolio are due entirely to gain or losses in the underlying securities i.e. the cash account and stock and not due to change in holding xt, and yt.

    Returning to our derivation, we can equate times in Eq.(2.18), to obtain

    yt=uS, (2.22)
    rxtBt=uS+12δ2S22(u)S2,

    If we set

    C0=P0,

    that C and P are call option and put option in finance, the interval value of strategy, then it must be the case that

    Ct=Pt,

    since call option or put option have the same dynamics this is true by construction after we equate terms in Eq.(2.19), with the corresponding terms in Eq.(2.22).

    Substituting Eq.(2.22) into Eq.(2.19), we obtain the Black-Scholes PDE

    rStuS+ut+12δ2S22(u)S2ru=0. (2.23)

    In order to solve Eq.(2.23), boundary conditions must be provided in the case of call option or put option.

    By applying system of Eq.(1.1) in Eq.(2.17), for construction of PDE model we need part of deterministic section of a stochastic differential equation model as,

    du(S,v,r,t)=ut(S,v,r,t)dt+us(S,v,r,t)d(S(t))+ur(S,v,r,t)d(r(t))+uv(S,v,r,t)d(v(t))+12uss(S,v,r,t)d(S(t))d(S(t))+12urr(S,v,r,t)d(r(t))d(r(t))+12uvv(S,v,r,t)d(v(t))d(v(t))+usv(S,v,r,t)d(S(t))d(v(t))+usr(S,v,r,t)d(S(t))d(r(t))+uvr(S,v,r,t)d(v(t))d(r(t)). (2.24)

    Heston-Cox-Ingersoll-Ross PDE equation [6,7,8,9],

    dudt=rSdudS+k(ηv)dudv+a(b(tT)r)dudr+12S2v2uS2+12δ12v2uv2 (2.25)
    +12δ22r2ur2+ρ12δ1Sv2uSv+ρ13Svr2uSr+ρ23δ1δ2vr2uvrru.

    By Eq.(2.25) we are obtained PDE model of the HCIR equation.

    In this study Legendre wavelet well-addressed in [5,10,18], are employed.The wavelet basis is constructed from a single function, which is called the mother wavelet. These basis functions are called wavelets and they are an orthonormal set. One of the most important wavelets is Legendre wavelets. The Legendre wavelets are obtained from Legendre polynomials. In the past decade special attention has been given to applications of wavelets. The main characteristic of Legendre wavelet is that it reduces to a system of algebraic equation.

    The function ψ(x)L2(R) is a mother wavelet and ψ(x)uv=|u|12ψ(xvu) which u,vR and u0 is a family of continuous wavelets. If we choose the dilation parameter u=an and the translation parameterv=manb where a>1,b>0,n and m are positive integer, we have the discrete orthogonal wavelets set.

    {ψ(x)n,m=|a|n2ψ(anxmb):m,nZ}

    The Legendre wavelet is constructed from Legendre function [4,15]. One dimension Legendre wavelets over the interval [0,1] are defined as,

    ψm,n(t)=(m+12)2k2Pm(2kx2n+1),ˆn12k10,o.w (3.1)

    With n=1,2...,2k1,m=0,1,2,...,M1. In Eq.(3.1) Pm are ordinary Legendre functions of order defined over the interval [-1, 1]. Legendre wavelet is an orthonormal set as,

    10ψ(x)n,mψ(x)n,mdx=δn,nδm,m, (3.2)

    The function f(x)L2([0,1]×[0,1]), may be expanded as the following,

    f(x)2k1n=1M1m=0cn,mψ(x)n,m, (3.3)

    Now we defined four dimensions Legendre wavelets in L4([0,1]×[0,1]×[0,1]×[0,1]), as the form

    ψn1,2,3,4,m1,2,3,4(x)={Am1,2,3,4(4i=1Pmi(2kixi2ni+1)),ni12ki1xini2ni1,i=1,..,40,o.w (3.4)

    that

    Am1,2,3,4=4i=1(mi+12)24i=1ki2,mi=0,1,...,Mi1,ni=1,2,...,2k11,i=1,..,4

    Four dimensions Legendre wavelets are orthonormal set over [0,1]×[0,1]×[0,1]×[0,1].

    10101010ψn1,2,3,4,m1,2,3,4(x,y,z,q)ψn1,2,3,4,m1,2,3,4(x,y,z,q)dxdydzdq= (3.5)
    δn1,n1,δn2,n2,δn3,n3,δn4,n4,δm1,m1,δm2,m2,δm3,m3,δm4,m4.

    The function u(x,y,z,q)L4([0,1]×[0,1]×[0,1]×[0,1]), can be expanded as fallows

    u(x,y,z,q)=X(x)Y(y)Z(z)Q(q)4i=1(2ki1ni=12Mi1mi=1cni,miψni,mi(x,y,z,q)). (3.6)

    Where

    cn1,2,3,4,m1,2,3,4=10101010X(x)Y(y)Z(z)Q(q)ψn1,2,3,4,m1,2,3,4dxdydzdq.

    The truncated version of Eq.(3.6), can be expressed as the following,

    u(x,y,z,q)=CTΨ(x,y,z,q), (3.7)

    Where C and Ψ(x,y,z,q), are coefficients matrix and wavelets vector matrix, respectively. The dimensions of those are {4i=1(2ki1Mi)}×1, and are given as the form

    C=[c10101010,...,c1010101M11,c10101020,...,c1010102M21,c10101030,...,c1010103M31,
    c1010102k110,...,c1010102k11M11,...,c1110101M11,...,c1120,...,c112M11,
    ...,c2k11M112k2102k3102k410,...,c2k11M112k21M212k31M312k41M41,...]T,

    and also

    Ψ=[ψ10101010,...,ψ1010101M11,ψ10101020,...,ψ1010102M21,ψ10101030,...,ψ1010103M31,
    ψ1010102k110,...,ψ1010102k11M11,...,ψ1110101M11,...,ψ1120,...,ψ112M11,
    ...,ψ2k11M112k2102k3102k410,...,ψ2k11M112k21M212k31M312k41M41,...]T.

    The integration of the product of four Legendre wavelet function vectors is obtained as

    10101010Ψ(x,y,z,q)ΨT(x,y,z,q)=I. (3.8)

    Where I, is diagonal unit matrix.(All of procedures in section 3, are established)

    Lemma 3.1: if hC([0,T],L2[0,1]) then the functions Cn,m(t) are continuous on [0,1] [13].

    Lemma 3.2: if hC1((0,T),L2[0,1]) then the function Cn,m(t) coefficient belong to C1(0,T), furthermore if htL2((0,T),L2[0,1]) then,

    dCn,m(t)dt=10h(t,x)tψn,m(x)dx. (3.9)

    Proof: see [13].

    The integration matrix for one variable x, define as follows

    x0Ψ(x,y,z,q)dx=PxΨ(x,y,z,q). (3.10)

    In which

    Px=1M12k1[ LFFF0LFF000L],

    Px, is a 2k112k212k312k41M1M2M3M4×2k112k212k312k41M1M2M3M4 matrix and L,F and O, are 2k11×M1×2k21×M2×2k31×M3×2k41×M4 matrices [14]. (In case of two variables Eq.(3.10), is established clearly).

    Theorem 3.1. A function f(x,y,z,l)L4([0,1)×[0,1)×[0,1)×[0,1)) is with bounded second derivative say |f(x,y,z,l)M, can be expanded as an infinite sum of Legendre wavelets and the series converges uniformly to the function f(x,y,z,l) that is [15],

    f(x,y,z,l)=n1=1m1=0n2=1m2=0n3=1m3=0n4=1m4=0cn1,..,4,m1,..,4ψn1,..,4,m1,..,4(x,y,z,l), (3.11)

    Proof:

    we have proved this theorem for multi-dimension states as following,

    cn1,..,4,m1,..,4=10101010f(x,y,z,l)dxdydzdl=
    ^n1+12k^n112k^n2+12k^n212k^n3+12k^n312k^n4+12k^n412kf(x,y,z,l)(2m1+12)12(2m2+12)12(2m3+12)12(2m4+12)12
    2k1+k2+k3+k42P(2k1xˆn)P(2k2xˆn)P(2k3xˆn)P(2k4xˆn)dxdydzdl.

    Now let,

    2k1xˆn=t1,2k2xˆn=t2,2k3xˆn=t3,2k4xˆn=t4,

    then,

    dx=12k1dt1,dy=12k2dt2,dz=12k3dt3,dl=12k4dt4.
    cn1,..,4,m1,..,4=11111111f(^n1+t12k1)f(^n2+t22k2)f(^n3+t32k3)f(^n4+t42k4)(2m1+12)12
    (2m2+12)12(2m3+12)12(2m4+12)122k1+k2+k3+k42P(t1)P(t2)P(t3)P(t4)dxdydzdl2k1+k2+k3+k4=

    with get out the constants as followes

    (2m1+12k1+1)12(2m2+12k2+1)12(2m3+12k3+1)12(2m4+12k4+1)1211111111f(^n1+t12k1)
    f(^n2+t22k2)f(^n3+t32k3)f(^n4+t42k4)Pm1(t1)Pm2(t2)Pm3(t3)Pm4(t4)dt1dt2dt3dt4=

    and polynomials are substituted

    (12k1+1(2m1+1))12(12k2+1(2m2+1))12(12k3+1(2m3+1))12(12k4+1(2m4+1))12
    11111111f(^n1+t12k1)f(^n2+t22k2)f(^n3+t32k3)f(^n4+t42k4)d(Pm1+1(t1)Pm11(t1))
    d(Pm2+1(t2)Pm21(t2))d(Pm3+1(t3)Pm31(t3))d(Pm4+1(t4)Pm41(t4))=
    (2m1+12k1+1)12(2m2+12k2+1)12(2m3+12k3+1)12(2m4+12k4+1)12f(^n1+t12k1)f(^n2+t22k2)f(^n3+t32k3)f(^n4+t42k4)
    (Pm1+1(t1)Pm11(t1))(Pm2+1(t2)Pm21(t2))(Pm3+1(t3)Pm31(t3))
    (Pm4+1(t4)Pm41(t4))|11|11|11|11=(123k1+1(2m1+1))12

    Constants in terms of polynomials are calculated with the others,

    (123k2+1(2m2+1))12(123k3+1(2m3+1))12(123k4+1(2m4+1))1211111111f(^n1+t12k1)
    f(^n2+t22k2)f(^n3+t12k3)f(^n4+t42k4)d(Pm1+2(t1)Pm1(t1)2m1+3Pm1(t1)Pm12(t1)2m11)
    d(Pm2+2(t2)Pm2(t2)2m2+3Pm2(t2)Pm22(t2)2m21)d(Pm3+2(t3)Pm3(t3)2m3+3Pm3(t3)Pm32(t3)2m31)
    d(Pm4+2(t4)Pm4(t4)2m4+3Pm4(t4)Pm42(t4)2m41)

    that,

    =|11111111f(^n1+t12k1)f(^n2+t22k2)f(^n3+t12k3)f(^n4+t42k4)
    (Pm1+2(t1)Pm1(t1)2m1+3Pm1(t1)Pm12(t1)2m11)(Pm2+2(t2)Pm2(t2)2m2+3Pm2(t2)Pm22(t2)2m21)
    (Pm3+2(t3)Pm3(t3)2m3+3Pm3(t3)Pm32(t3)2m31)(Pm4+2(t4)Pm4(t4)2m4+3Pm4(t4)Pm42(t4)2m41)|2
    |11111111f(^n1+t12k1)f(^n2+t22k2)f(^n3+t12k3)f(^n4+t42k4)
    ((2m11)Pm1+2(t1)(4m1+2)Pm1(t1)+(2m1+3)2m1+3)...

    Using integration properties as following,

    ((2m41)Pm4+2(t4)(4m4+2)Pm4(t4)+(2m1+3)2m4+3)dt1dt2dt3dt4|2
    |11111111f(^n1+t12k1)f(^n2+t22k2)f(^n3+t12k3)f(^n4+t42k4)dt1dt2dt3dt4|2×
    11111111|((2m11)Pm1+2(t1)(4m1+2)Pm1(t1)+(2m1+3)2m1+3)...dt1dt2dt3dt4|2
    2M22M22M22M211111111((2m11)Pm1+2(t1)(4m1+2)Pm1(t1)+(2m1+3)2m1+3)
    ...dt1dt2dt3dt4=

    Upper bounds are noted as,

    16M16(2m1+3)2(2m11)2...(2m4+3)2(2m41)2((2m11)222m1+5+
    (4m1+2)222m1+1+(2m1+3)222m13)...((2m41)222m4+5+
    (4m4+2)222m4+1+(2m4+3)222m43)<

    Now (2m11)2<(2m1+3)2, and (2m1+1)2<(2m1+3)2, then

    =331776M16(2m11)((2m13)...(2m41)(2m43))2.

    Therefore we have,

    |cn1,..,4,m1,..,4|<(576)121(2n)121(2m13)2...(2m43)2.

    Thus the series n1=1m1=0n2=1m2=0n3=1m3=0n4=1m4=0cn1,..,4,m1,..,4, is ab-solute convergent, it follows that

    n1=1m1=0n2=1m2=0n3=1m3=0n4=1m4=0cn1,..,4,m1,..,4ψn1,..,4,m1,..,4(x,y,z,l),

    convergence to the function f(x,y,z,l), uniformly.

    Notice: This result for two dimensions in theorem (3.1), is calculated as upper bound,

    |cn1n2m1m2|<(288)1212n521(2m13)2(2m23)2.

    In this methodology section, the Black-Scholes model for American put problem have used in the form of moving-boundary problem.

    Lemma 4.1: For American options, we have [5,6]

    ⅰ) The optimal exercise time for American call option is T and we have

    C=c,

    ⅱ) The optimal exercise time for American put option as earlier as possible, i.e. t and we have

    Pp,

    ⅲ) The put-call parity for American option:

    SK<CP<Sket(Tt),

    As a consequence,

    PK.

    Letting S be the price of an asset A at time t the American early exercise constraint leads to the following model for the value P(S,t) of an American put option to sell the asset A.

    Pt+12δ2S22PS2+rSPSrP=0,0t<T (4.1)
    P(S,T)=max(ES,0),S0,
    PS(¯S,t)=1,
    P(S,t)=ES,0S<S(¯t).

    Where S(¯t) represent the free boundary. The parameters δ,r and E represent the volatility of the underlying asset, the interest rate, and the exercise price of the option. And the partial differential equation (PDE) associated with Eq.(2.25) for the fair values of European- style options forms a time- dependent connection- diffusion- reaction equation with mixed spatial- derivative terms.

    ut=12S2v2uS2+12δ21v2uv2+12δ22r2ur2+ρ12δ1Sv2uSv+ρ13δ2Svr2uSr (4.2)
    ρ23δ1δ2vr2uvr+rSuS+k(ηv)uv+a(b(Tt)r)urru.

    On the unbounded three- dimensional spatial domain S>0,v>0,r>0 with 0<tT. u(S,v,r,t) denotes the fair value of a European- style option if at the time τ=Tt the asset price, its variance, and the interest rate be equal to S,v and r respectively. Eq.(4.2) is known as the Heston- Cox- Ingersoll- Ross (HCIR) PDE. This equation usually appears with initial and boundary conditions which are determined by the specific option under consideration. Initial conditions in European call options equation is usually as follows,

    u(S,v,r,T)=max(0,Sk). (4.3)

    With a given strike price k>0.

    In this methodology section the Eq.(4.1) and Eq.(4.2), have been solved using two dimensional and four dimensional Legendre wavelets respectively,

    First we assume for B-S PDE,

    Pt=α2PS2αPS+αP, (4.4)

    With boundary condition P(S,0)=β1(S,t), and P(S,0)t=β2(S,t).

    Those coefficients of αi the same parameters and variables which be in Eq.(4.1), in practice they will be calculate with this technique in follow.

    Pt=CT1(S,t)Ψ(S,t), (4.5)

    Integrating (4.4) with respect to second variable over [0,t], we get

    P(S,t)=CT1(S,t)PΨ(S,t)+β2(S,t), (4.6)

    And also

    PS=dCT(S,t)dSPΨ(S,t)+dβ2(S,t)dS, (4.7)
    2PS2=d2CT(S,t)dS2PΨ(S,t)+d2β2(S,t)dS2,

    We have also

    1=dTΨ(x),x=eTΨ(x). (4.8)

    Substituting Eqs.(4.4) to (4.8) in Eq.(4.1), we obtain

    CT1(S,t)Ψ=α1(d2CT(S,t)dS2PΨ+d2β2(S,t)dS2dTΨ)α2(dCT(S,t)dSPΨ+dβ2(S,t)dSdTΨ) (4.9)
    +α3(CT1(S,t)PΨ+β2dTΨ).

    The solution of B-S PDE, P(S,t) can be calculate when Eq.(4.9) can be solved For unknown coefficients of the vector C(S,t).

    C=[0.121771121770.2214022140e1t,0.1661746162+0.3451318952t,0.7030464532e10.1278266279e1t,0.9594095941e1+0.1992619926t]T.

    Let for HCIR PDE in Eq.(4.2), we assume as following,

    ut=α12uS2+α22uv2+α32ur2+α42uSv+α52uSr+α62uvr+α7uS (4.10)
    +α8uv+α9ur+α10u.

    Also with boundary condition

    u(S,v,r,t)=β1(S,v,r,t),

    and

    u(S,v,r,0)t=β2(S,v,r,t).

    Parameters of Eq.(4.2), by using method as following,

    ut=CT1(S,v,r,t)Ψ(S,v,r,t), (4.11)

    Integrating right hand side in Eq.(4.10), with respect to differential equations over [0, t] we get As well as

    uS=dCT(S,v,r,t)dSPΨ(S,v,r,t)+dβ2(S,v,r,t)dS, (4.12)
    uv=dCT(S,v,r,t)dvPΨ(S,v,r,t)+dβ2(S,v,r,t)dv,
    ur=dCT(S,v,r,t)drPΨ(S,v,r,t)+dβ2(S,v,r,t)dr.

    We have

    2uS2=d2CT(S,v,r,t)dS2PΨ(S,v,r,t)+d2β2(S,v,r,t)dS2, (4.13)
    2uv2=d2CT(S,v,r,t)dv2PΨ(S,v,r,t)+d2β2(S,v,r,t)dv2,
    2ur2=d2CT(S,v,r,t)dr2PΨ(S,v,r,t)+d2β2(S,v,r,t)dr2.

    And also

    2uSv=ddSdCT(S,v,r,t)dvPΨ(S,v,r,t)+ddSdβ2(S,v,r,t)dv, (4.14)
    2uSr=ddSdCT(S,v,r,t)drPΨ(S,v,r,t)+ddSdβ2(S,v,r,t)dr,
    2uvr=ddvdCT(S,v,r,t)drPΨ(S,v,r,t)+ddvdβ2(S,v,r,t)dr.

    Substituting (4.11) to (4.14) in (48), we obtain

    CT(S,v,r,t)Ψ=α1(2CT(S,v,r,t)S2PΨ+d2β2(S,v,r,t)dS2dTΨ) (4.15)
    +α2(2CT(S,v,r,t)v2PΨ+d2β2(S,v,r,t)dv2dTΨ)+α3(2CT(S,v,r,t)r2PΨ+d2β2(S,v,r,t)dr2dTΨ)
    +α4(ddSdCT(S,v,r,t)dvPΨ+ddSdβ2(S,v,r,t)dvdTΨ)+α5(ddSdCT(S,v,r,t)drPΨ+ddSdβ2(S,v,r,t)drdTΨ)
    +α6(ddvdCT(S,v,r,t)drPΨ+ddvdβ2(S,v,r,t)drdTΨ)+α7(CT(S,v,r,t)SPΨ+dβ2(S,v,r,t)dSdTΨ)
    +α8(CT(S,v,r,t)vPΨ+dβ2(S,v,r,t)dvdTΨ)+α9(CT(S,v,r,t)rPΨ+dβ2(S,v,r,t)drdTΨ)
    α10(CT(S,v,r,t)PΨ+β2dTΨ).

    This system can be solved for unknown coefficients, which are calculated,

    C[1,1]=0.1816780437e2×sqrt(v)×s×t+.23100890490.3191757882e2×s×t20.2243167683e2×sqrt(v)×s0.3663030704×s×t+0.5319596471e3×t2+0.4571378943×s+0.8386506682e2×t,
    C[1,2]=0.4720134035e3×sqrt(v)×s×t+0.8292430227e3×s×t2+0.5827920594e3×sqrt(v)×s+0.9516832935e1×s×t0.1382071705e3×t20.1187679088×s0.2178878351e2×t+0.1997897471,
    C[1,3]=0,C[1,4]=0,C[1,5]=0.2097293811,C[1,6]=0.1705107084,C[1,7]=0,C[1,8]=0,C[1,9]=0.04660652914,C[1,10]=0.3789126853e1,C[1,11]=0,C[1,12]=0,C[1,13]=0.4036243822e1,C[1,14]=0.3281480113e1,C[1,15]=0,C[1,16]=0.

    In this section, values of parameters are used in B-S PDE model (Figure 1): k1=k2=1,r=0.2,E=0.2,δ=0.001, and HCIR PDE model (Figures 2 and 3): k=0.1,η=0.12,ρ12=0.6,ρ13=0.2,ρ23=0.4,δ1=0.04,δ2=0.03, respectively. Appling values of parameters in models, numerical results are calculated. Due to the accuracy and convergence of spectral method which has been reviewed in [0,1], the numerical results obtained from this method can be assured. Therefore it can be used for future scams to compare with the same methods.

    Figure 1.  Figure of solution B-S PDE model by multi-LWM.
    Figure 2.  Figure of solution HCIR PDE model by multi-LWM.
    Figure 3.  Figure of solution HCIR PDE model by ADM.

    By choosing M=2,4 and k=1 in Eq.(2.23) and M=1,2 in Eq.(2.25) the models are solved from the coefficients vector of C that was found. Then the results were compared by Adomian decomposition method (ADM), as a semi-analytical method [20,21]. Tables 1 and 2 give values of pricing options solutions in B-S and HCIR models.

    Table 1.  Corresponding solution of Black-Scholes PDE by LWM, ADM.
    t P(LWM,M=2) P(LWM,M=4) P(ADM) Error
    0.1 0.3060753725 0.3060954392 0.3068021827 0.000726810
    0.2 0.02615249781 0.02664386902 0.02613346987 0.000019028
    0.4 0.04431200702 0.04942003420 0.04430950409 0.000005503
    0.6 0.05847852762 0.05882487996 0.05862488320 0.000146355
    0.8 0.08565205963 0.08534345864 0.08534345864 0.000308601
    1 0.05083260299 0.0509281709 0.05075733333 0.000075268

     | Show Table
    DownLoad: CSV
    Table 2.  Corresponding solution of Heston-Cox-Ingersoll-Ross PDE by LWM, ADM.
    (t,s) U(LWM,M=1) U(LWM,M=2) U(ADM) Error
    (5,50) 0.1950128206e+3 0.1955141009e+3 0.1951180661e+3 0.0003960348
    (10,100) 0.2721204331e+4 0.2725170985e+4 0.2789681554e+4 0.0064510569
    (12,150) 0.6421235221e+4 0.6430612189e+4 0.6448380051e+4 0.0017767862
    (15,200) 0.1461103897e+5 0.1470153697e+5 0.1486797294e+5 0.0016643597
    (17,250) 0.2471265331e+5 0.2474067533e+5 0.2471107229e+5 0.0002960304
    (19,300) 0.3852234601e+5 0.3858603408e+5 0.3807240307e+5 0.0051363101
    (22,380) 0.6881012512e+5 0.6890201411e+5 0.6860384740e+5 0.0029816671

     | Show Table
    DownLoad: CSV

    The multi-dimensional Legendre wavelet method has been applied to solve the Black-Scholes and Heston-Cox-Ingersoll-Ross models which these PDE equations have been derived from stochastic differential equations by using one of the important stochastic calculous Lemma named Ito. Generally the present method (LWM) reduces the partial differential equation into a set of algebraic equation and it can be useful in high degree derivatives problems, therefore considering the properties and using structure of LWM leads to save the time. Multi- dimensional Legendre wavelet is orthonormal set and its integration of the product is a diagonal matrix as in the case of the one dimension Legendre wavelet. In addition the Theorem 3.1 has been proved in paper which it could be determine an upper bound for coefficients of vector C. Also in this work results of LWM have been compared with the Adomian Decomposition Method (ADM) as a semi-analytical method and they have been shown for different parameters in Tables 12 and Figures 13, finally.

    The authors declare that there is no conflicts of interest in this paper.



    [1] G. Bao, T. Yin, Recent progress on the study of direct and inverse elastic scattering problems (in Chinese), Sci. Sin. Math., 47 (2017), 1103–1118. https://doi.org/10.1360/N012016-00198 doi: 10.1360/N012016-00198
    [2] D. Colton, R. Kress, Inverse Acoustic and Electromagnetic Scattering Theory, Springer Nature, Berlin, 2019. https://doi.org/10.1007/978-3-030-30351-8
    [3] F. Cakoni, D. Colton, A Qualitative Approach to Inverse Scattering Theory, Springer, US, 2014. https://doi.org/10.1007/978-1-4614-8827-9
    [4] F. Cakoni, D. Colton, Qualitative Methods in Inverse Scattering Theory, Springer, Vienna, 2006. https://doi.org/10.1007/3-540-31230-7
    [5] H. Ammari, E. Bretin, J. Garnier, H. Kang, H. Lee, A. Wahab, Mathematical Methods in Elasticity Imaging, Princeton University Press, New Jersey, 2015.
    [6] J. H. Bramble, J. E. Pasciak, A note on the existence and uniqueness of solutions of frequency domain elastic wave problems: a priori estimates in H1, J. Math. Anal. Appl., 345 (2008), 396–404. https://doi.org/10.1016/j.jmaa.2008.04.028 doi: 10.1016/j.jmaa.2008.04.028
    [7] G. Bao, G. Hu, J. Sun, T. Yin, Direct and inverse elastic scattering from anisotropic media, J. Math. Pures Appl., 117 (2018), 263–301. https://doi.org/10.1016/j.matpur.2018.01.007 doi: 10.1016/j.matpur.2018.01.007
    [8] H. Liu, M. Petrini, L. Rondi, J. Xiao, Stable determination of sound-hard polyhedral scatterers by a minimal number of scattering measurements, J. Differ. Equations, 262 (2017), 1631–1670. https://doi.org/10.1016/j.jde.2016.10.021 doi: 10.1016/j.jde.2016.10.021
    [9] J. Li, H. Liu, Y. Wang, Recovering an electromagnetic obstacle by a few phaseless backscattering measurements, Inverse Probl., 333 (2017), 035011. https://doi.org/10.1088/1361-6420/aa5bf3 doi: 10.1088/1361-6420/aa5bf3
    [10] J. Li, H. Liu, Q. Wang, Enhanced multilevel linear sampling methods for inverse scattering problems, J. Comput. Phys., 257 (2014), 554–571. https://doi.org/10.1016/j.jcp.2013.09.048 doi: 10.1016/j.jcp.2013.09.048
    [11] T. Arens, Linear sampling methods for 2D inverse elastic wave scattering, Inverse Probl., 17 (2001), 1445–1464. https://doi.org/10.1088/0266-5611/17/5/314 doi: 10.1088/0266-5611/17/5/314
    [12] A. Charalambopoulos, A. Kirsch, K. Anagnostopoulos, D. Gintides, K. Kiriaki, The factorization method in inverse elastic scattering from penetrable bodies, Inverse Probl., 23 (2007), 27–51. https://doi.org/10.1088/0266-5611/23/1/002 doi: 10.1088/0266-5611/23/1/002
    [13] G. Hu, A. Kirsch, M. Sini, Some inverse problems arising from elastic scattering by rigid obstacles, Inverse Probl., 29 (2013), 015009. https://doi.org/10.1088/0266-5611/29/1/015009 doi: 10.1088/0266-5611/29/1/015009
    [14] Z. Chen, G. Huang, Reverse time migration for extended obstacles: elastic waves (in Chinese), Sci. Sin. Math., 45 (2015), 1103–1114. https://doi.org/10.1360/N012014-00097 doi: 10.1360/N012014-00097
    [15] P. Li, Y. Wang, Z. Wang, Y. Zhao, Inverse obstacle scattering for elastic waves, Inverse Probl., 32 (2016), 115018. https://doi.org/10.1088/0266-5611/32/11/115018 doi: 10.1088/0266-5611/32/11/115018
    [16] J. Antˊonio, A. Tadeu, L. Godinho, A three-dimensional acoustics model using the method of fundamental solutions, Eng. Anal. Boundary Elem., 32 (2008), 525–531. https://doi.org/10.1016/j.enganabound.2007.10.008 doi: 10.1016/j.enganabound.2007.10.008
    [17] A. Karageorghis, B. T. Johansson, D. Lesnic, The method of fundamental solutions for the identification of a sound-soft obstacle in inverse acoustic scattering, Appl. Numer. Math., 62 (2012), 1767–1780. https://doi.org/10.1016/j.apnum.2012.05.011 doi: 10.1016/j.apnum.2012.05.011
    [18] Y. Sun, X. Lu, B. Chen, The method of fundamental solutions for the high frequency acoustic-elastic problem and its relationship to a pure acoustic problem, Eng. Anal. Boundary Elem., 156 (2023), 299–310. https://doi.org/10.1016/j.enganabound.2023.08.010 doi: 10.1016/j.enganabound.2023.08.010
    [19] A. Karageorghis, D. Lesnic, L. Marin, The MFS for the identification of a sound-soft interior acoustic scatterer, Eng. Anal. Boundary Elem., 83 (2017), 107–112. https://doi.org/10.1016/j.enganabound.2017.07.021 doi: 10.1016/j.enganabound.2017.07.021
    [20] A. Karageorghis, D. Lesnic, L. Marin, The method of fundamental solutions for the identification of a scatterer with impedance boundary condition in interior inverse acoustic scattering, Eng. Anal. Boundary Elem., 92 (2018), 218–224. https://doi.org/10.1016/j.enganabound.2017.07.005 doi: 10.1016/j.enganabound.2017.07.005
    [21] D. Colton, R. Kress, Using fundamental solutions in inverse scattering, Inverse Probl., 22 (2006), R49. https://doi.org/10.1088/0266-5611/22/3/R01 doi: 10.1088/0266-5611/22/3/R01
    [22] C. J. S. Alves, N. F. M. Martins, S. S. Valtchev, Extending the method of fundamental solutions to non-homogeneous elastic wave problems, Appl. Numer. Math., 115 (2017), 299–313. https://doi.org/10.1016/j.apnum.2016.06.002 doi: 10.1016/j.apnum.2016.06.002
    [23] W. Yin, W. Yang, H. Liu, A neural network scheme for recovering scattering obstacles with limited phaseless far-field data, J. Comput. Phys., 417 (2020), 109594. https://doi.org/10.1016/j.jcp.2020.109594 doi: 10.1016/j.jcp.2020.109594
    [24] Z. Wei, X. Chen, Deep-learning schemes for full-wave nonliner inverse scattering problems, IEEE Trans. Geosci. Remote Sens., 57 (2019), 1849–1860. https://doi.org/10.1109/TGRS.2018.2869221 doi: 10.1109/TGRS.2018.2869221
    [25] E. Chung, J. G. Park, Sentence-chain based Seq2seq model for corpus expansion, ETRI J., 39 (2017), 455–466. https://doi.org/10.4218/etrij.17.0116.0074 doi: 10.4218/etrij.17.0116.0074
    [26] J. Torres, C. Vaca, L. Terán, C. L. Abad, Seq2Seq models for recommending short text conversations, Expert Syst. Appl., 150 (2020), 113270. https://doi.org/10.1016/j.eswa.2020.113270 doi: 10.1016/j.eswa.2020.113270
    [27] Y. Khoo, L. Ying, SwitchNet: a neural network model for forward and inverse scattering problems, SIAM J. Sci. Comput., 41 (2019), A3182–A3201. https://doi.org/10.1137/18M1222399 doi: 10.1137/18M1222399
    [28] G. Fairweather, A. Karageorghis, P. A. Martin, The method of fundamental solutions for scattering and radiation problems, Eng. Anal. Boundary Elem., 27 (2003), 759–769. https://doi.org/10.1016/S0955-7997(03)00017-1 doi: 10.1016/S0955-7997(03)00017-1
    [29] Y. Sun, F. Ma, A meshless method for the Cauchy problem in linear elastodynamics, Appl. Anal., 93 (2014), 2647–2667. https://doi.org/10.1080/00036811.2014.882913 doi: 10.1080/00036811.2014.882913
    [30] P. Zhang, P. Meng, W. Yin, H. Liu, A neural network method for time-dependent inverse source problem with limited-aperture data, J. Comput. Appl. Math., 421 (2023), 114842. https://doi.org/10.1016/j.cam.2022.114842 doi: 10.1016/j.cam.2022.114842
    [31] D. Xu, Z. Li, W. Wu, X. Ding, D. Qu, Convergence of gradient method descent algorithm for recurrent neuron, Int. Symp. Neural Network, 4493 (2007), 117–122. https://doi.org/10.1007/978-3-540-72395-0_16 doi: 10.1007/978-3-540-72395-0_16
    [32] D. Xu, Z. Li, W. Wu, Convergence of gradient method for fully recurrent neural network, Soft Comput., 14 (2010), 245–250. https://doi.org/10.1007/s00500-009-0398-0 doi: 10.1007/s00500-009-0398-0
  • This article has been cited by:

    1. Kazem Nouri, Milad Fahimi, Leila Torkzadeh, Dumitru Baleanu, Numerical method for pricing discretely monitored double barrier option by orthogonal projection method, 2021, 6, 2473-6988, 5750, 10.3934/math.2021339
    2. Lei Fu, Huanhe Dong, Chaudry Masood Khalique, Hongwei Yang, CONSERVATION LAWS, EXACT SOLUTIONS OF TIME-SPACE FRACTIONAL GENERALIZED GINZBURG-LANDAU EQUATION FOR SHALLOW WAKE FLOWS, 2021, 11, 2156-907X, 874, 10.11948/20200053
    3. Abdulaziz Alsenafi, Fares Alazemi, Javad Alavi, An efficient algorithm to solve the geometric Asian power option price PDE under the stochastic volatility model, 2024, 1017-1398, 10.1007/s11075-024-01794-z
  • Reader Comments
  • © 2023 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(1399) PDF downloads(59) Cited by(3)

Figures and Tables

Figures(9)  /  Tables(1)

Other Articles By Authors

/

DownLoad:  Full-Size Img  PowerPoint
Return
Return

Catalog