A Comparative Study on Haar Wavelet and Hosaya Polynomial for the numerical solution of Fredholm integral equations

Integral equations have motivated a large amount of research work in recent years. Integral equations find its applications in various fields of mathematics, science and technology has been studied extensively both at the theoretical and practical level. In particular, integral equations arise in fluid mechanics, biological models, solid state physics, kinetics in chemistry etc. In most of the cases, it is difficult to solve them, especially analytically [1]. Analytical solutions of integral equations, however, either does not exist or are difficult to find. It is precisely due to this fact that several numerical methods have been developed for finding solutions of integral equations.

Consider the Fredholm integral equation:

(1)

\begin{matrix} y (x) = f (x) + \int_{0}^{1} k (x, t) y (t) d t 0 \leq x, t \leq 1 \end{matrix}

$$\begin{eqnarray}\label{1} y(x)=f(x)+\int_{0}^{1}k(x,t) y(t)dt \,\,\,\,\,\, 0\leq x,t\leq 1 \end{eqnarray}$$

where f (x) and the kernels k(x,t) are assumed to be in L₂(R) on the interval 0 ≤ x,t ≤ 1. We assume that Eq.(1) has a unique solution y to be determined. There are several numerical methods for approximating the solutionof Fredholm integral equations are known and many different basic functions have been used. Such as, Lepik et al. [2] applied the Haar Wavelets. Maleknejad et. al [3] applied a combination of Hybrid taylor and block-pulse functions, Rationalized haar wavelet [4], Hermite Cubic splines [5]. Muthuvalu et al. [6] applied Half-sweep arithmetic mean method with composite trapezoidal scheme. In previous paper [35], we compared with [5] for the numerical solution of Fredholm integral equations.

Wavelets theory is a relatively new and an emerging tool in applied mathematical research area. It has been applied in a wide range of engineering disciplines; particularly, signal analysis for waveform representation and segmentations, time-frequency analysis and fast algorithms for easy implementation. Wavelets permit the accurate representation of a variety of functions and operators. Moreover, wavelets establish a connection with fast numerical algorithms [7, 8]. Since 1991 the various types of wavelet method have been applied for the numerical solution of different kinds of integral equations [2]. In the previous work, system analysis via Haar wavelets was led by Chen and Hsiao [9], who first derived a Haar operational matrix for the integrals of the Haar function vector and put the applications for the Haar analysis into the dynamic systems. Recently, Haar wavelet method is applied for different type of problems. Namely, Bujurke et al. [10, 11, 12] used the single term Haar wavelet series for the numerical solution of stiff systems from nonlinear dynamics, nonlinear oscillator equations and Sturm-Liouville problems. Shiralashetti et al. [13 14 15,16, 18] applied for the numerical solution of Klein?Gordan equations, multi-term fractional differential equations, singular initial value problems,nonlinear Fredholm integral equations, Riccati and Fractional Riccati Differential Equations. Shiralashetti et al. [17] have introduced the adaptive gird Haar wavelet collocation method for the numerical solution of parabolic partial differential equations.

At present, field of a graph theory is a dynamic field in both theory and applications. Graphs can be utilized as a modeling tool to solve a many problems of practical importance. Structural graph theory is naturally a very wide field itself, and the current state of art is more advanced in some parts than in others.

A graph G consists of a finite nonempty set V of n points(vertex) together with a prescribed set X of m unordered pairs of distinct points of V. Each pair x = u,v of points in X is a edge of G, and x is said to join u and v. We write x = uv and say that u and v are adjacent points (sometimes denoted u adj v).

Let G be a simple connected graph with a vertex set V(G) = {_v1,v₂, · · · ,v_n} and edge set E(G) = {e1,e₂, · · · ,e_n}, where |V(G)| = n and |E(G)| = m. The degree of a vertex v in G is the number of edges incident to it and denoted by deg_G(v). The distance between the vertices v_i and v _j in G is equal to the length of the shortest path joining them and is denoted by d (v_i,v_j|G).

The Wiener index W =W(G) of a graph G is defined as the sum of the distances between all pairs of vertices of G, that is

W = W (G) = \sum_{1 \leq i < j \leq n} d (u_{i}, v_{j} | G) .

$$W= W(G) = \sum\limits_{1 \leq i < j \leq n} d(u_i, v_j | G) .$$

This index was putforward by Wiener [21] in 1947 for approximating the boiling points of alkanes. The effect of approximation was surprisingly good. From that point forward, the Wiener index has attracted the attention of chemists.

Recently several modifications of the Wiener index were put forward, of which we mention the following: the hyper–Wiener index [22, 23]:

(2)

W W = W W (G) = \frac{1}{2} \sum_{u < v} (d (u, v)^{2} + d (u v))

$$\begin{equation}\label{hy} WW = WW(G) = {1 \over 2}\,\sum_{u<v} \left[ d(u,v)^2 + d(uv) \right] \end{equation}$$

the Harary index [24]:

(3)

H (G) = \sum_{u < v} \frac{1}{d (u, v)}

$$\begin{equation}H(G) = \sum_{u<v} \frac{1}{d(u,v)} \end{equation}$$

It is worth noting that all the above structure–descriptors are either special cases of, or are simply related to the graph invariant W_l , defined as [25, 26]

(4)

W_{λ} = W_{λ} (G) = \sum_{k \geq 1} d (G, k) k^{λ}

$$\begin{equation}\label{w} W_\lambda = W_\lambda(G) = \sum_{k \geq 1} d(G,k)\,k^\lambda \end{equation}$$

where d (G,k) is the number of pairs of vertices of the graph G whose distance is k , and where λ is some real (or complex) number. Evidently,

\begin{matrix} W & = & W_{1} \\ W W & = & \frac{1}{2} W_{2} + \frac{1}{2} W_{1} \\ H a & = & W_{- 2} \\ R W & = & W_{- 1} . \end{matrix}

$$\begin{eqnarray*} W & = & W_1 \\ WW & = & {1 \over 2}\,W_2 + {1 \over 2}\,W_1 \\ Ha & = & W_{-2} \\ RW & = & W_{-1} . \end{eqnarray*}$$

Another related quantity is the Hosoya polynomial. In the present work, a comparison of a Haar wavelet and Hosoya polynomial methods for the numerical solution of Fredholm integral equations is proposed.

Properties of Haar wavelet and Hosoya polynomial

2.1

Haar wavelet

The scaling function h₁(x) for the family of the Haar wavelets is defined as

(5)

h_{1} (x) = (\begin{matrix} 1 & f o r x \in [0, 1) \\ 0 & o t h e r w i s e \end{matrix})

$$\begin{equation}h_{1}(x) = \left\{ \begin{array}{ll} 1 & for\hspace{0.2cm} x\in[0, 1) \\ 0 & otherwise \\ \end{array} \right. \end{equation}$$

The Haar wavelet family for x 2 [0,1) is defined as,

(6)

h_{i} (x) = (\begin{matrix} 1 & f o r x \in (α, β) \\ - 1 & f o r x \in (β, γ) \\ 0 & o t h e r w i s e \end{matrix})

$$\begin{equation}h_{i}(x) = \left\{ \begin{array}{ll} 1 & for \hspace{0.2cm} x\in\left[\alpha ,\beta\right) \\ \\ -1 & for \hspace{0.2cm} x\in\left[ \beta, \gamma\right) \\ \\ 0 & otherwise \\ \end{array} \right. \end{equation}$$

where $α = \frac{k}{m}, β = \frac{k + 0.5}{m}, γ = \frac{k + 1}{m}$ $\alpha = {\dfrac{k}{m}}, \beta = {\dfrac{k+0.5}{m}}, \gamma = {\dfrac{k+1}{m}}$

In the above definition the integer, m = 2^l , l = 0,1,⋯,J, indicates the level of resolution and integer k = 0,1,⋯,m-1 is the translation parameter. Maximum level of resolution is J. The index i in (6) is calculated using, i = m +k +1: In case of minimal values m = 1 and k = 0 then i = 2: The maximal value of i is N = 2^J⁺¹.

Let us define the collocation points $x_{j} = \frac{j - 0.5}{K}, j = 1, 2, . . ., N,$ $x_{j}=\dfrac{j-0.5}{K}, j=1, 2, . . . , N,$discretize the Haar function h_i(x) and the corresponding Haar coefficient matrix H (i, j) = (h_i(x_j)), which has the dimension N × N.

If J = 3 )N = 16,

Eq. 6 generates the Haar matrix,

H (16, 16) = (\begin{matrix} 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & - 1 & - 1 & - 1 & - 1 & - 1 & - 1 & - 1 & - 1 \\ 1 & 1 & 1 & 1 & - 1 & - 1 & - 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 1 & 1 & - 1 & - 1 & - 1 & - 1 \\ 1 & 1 & - 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 1 & 1 & - 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & - 1 & - 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & - 1 & - 1 \\ 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & - 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & - 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & - 1 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & - 1 \end{matrix})

$$H(16,16) = \left(\begin{array}{cccccccccccccccc} 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1\\ 1 & 1 & 1 & 1 & 1 & 1 & 1 & 1 & -1 & -1 & -1 & -1 & -1 & -1 & -1 & -1\\ 1 & 1 & 1 & 1 & -1 & -1 & -1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & 1 & 1 & -1 & -1 & -1 & -1\\ 1 & 1 & -1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 1 & 1 & -1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & -1 & -1 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & 1 & -1 & -1\\ 1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 1 & -1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & -1 & 0 & 0 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & -1 & 0 & 0 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & -1 & 0 & 0\\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 & -1 \end{array} \right)$$

As Haar Wavelets are orthogonal; this mean that any square integrable function over [0,1] can be expanded into Haar wavelets series as:

(7)

\begin{matrix} f (x) = \sum_{i = 1}^{\infty} a_{i} h_{i} (x) \end{matrix}

$$\begin{eqnarray}f(x)=\sum_{i=1}^{\infty}a_ih_i(x)\end{eqnarray}$$

where a_i’s are Haar wavelet coefficients.

If f (x) be piecewise constant, then sum can be terminated to finite term, that is

(8)

f (x) = \sum_{i = 1}^{N} a_{i} h_{i} (x_{i}) = a^{T} H

$$\begin{equation}f(x)=\sum_{i=1}^{N}a_ih_i(x_i)=a^TH \\ \end{equation}$$

(9)

a^{T} = (a_{1}, a_{2}, . . ., a^{N}), H = (h_{1} (x), h_{2} (x), . . ., h_{N} (x))^{T} .

$$\begin{equation}a^T=(a_1,a_2,...,a^N), H=(h_1(x),h_2(x),. . ., h_N(x))^T. \end{equation}$$

2.2

Hosoya polynomial

The Hosoya polynomial of a graph is a generating function about distance distributing, introduced by Hosoya [27] in 1988. For a connected graph G, the Hosoya polynomial denoted by H (G,λ) is defined as

(10)

H (G, λ) = \sum_{k \geq 0} d (G, k) λ^{k}

$$\begin{equation}\label{h} H(G, \lambda) = \sum\limits_{k\geq 0} d(G, k)\lambda^{k} \end{equation}$$

where d (G,k) is the number of pairs of vertices of G that are at distance k and λ is the parameter.

Comparing (4) and (10) we see that both W_l and H(λ) are determined by the numbers d (G,k) , k = 1,2, . . . . Indeed, from knowing H (G,l) one can deduce W_l (G) and vice versa. Thus, from the Hosoya polynomial it is possible to calculate a large variety of distance–based molecular structure–descriptors.

The connection between the Hosoya polynomial and the Wiener index is elementary [27, 28]:

W (G) = H^{'} (G, 1)

$$W(G) = H'(G,1)$$

where H^'(G,λ) is the first derivative of H (G,l) . The hyper–Wiener index can be computed from the first and second derivatives as [29]

W W (G) = H^{'} (G, 1) + \frac{1}{2} H^{″} (G, 1) .

$$WW(G) = H'(G,1) + {1 \over 2}\,H''(G,1) .$$

The path on n vertices (or of length n -1) is the graph with n vertices-say, 1,2, .,n and with n -1 edges, such that vertices i and i+1 are adjacent, i = 1,2,...,n-1. For any positive integer n we denote path as P_n, then Hosoya polynomial of path is:

\begin{matrix} H (P_{1}, λ) & = & \sum_{k \geq 0} d (P_{1}, k) λ^{k} = 1 \\ H (P_{2}, λ) & = & \sum_{k \geq 0} d (P_{2}, k) λ^{k} = λ + 2 \\ H (P_{3}, λ) & = & \sum_{k \geq 0} d (P_{3}, k) λ^{k} = λ^{2} + 2 λ + 3 \\ . \\ . \\ . \\ H (P_{n}, λ) & = & \sum_{k \geq 0} d (P_{n}, k) λ^{k} = n + (n - 1) λ + (n - 2) λ^{2} + . . . + [n - (n - 2)] λ^{n - 2} + [n - (n - 1) λ^{n - 1}] \end{matrix}

$$\begin{eqnarray*} H(P_1, \lambda) & = & \sum\limits_{k\geq 0} d(P_1, k)\lambda^{k}=1 \\ H(P_2, \lambda) & = & \sum\limits_{k\geq 0} d(P_2, k)\lambda^{k}=\lambda+2 \\ H(P_3, \lambda) & = & \sum\limits_{k\geq 0} d(P_3, k)\lambda^{k}=\lambda^2+2\lambda+3 \\ .\\ .\\ .\\ H(P_n, \lambda) & = & \sum\limits_{k\geq 0} d(P_n, k)\lambda^{k}=n+(n-1)\lambda +(n-2) \lambda^2+ . . . + [n-(n-2)]\lambda^{n-2}+[n-(n-1)\lambda^{n-1}] \end{eqnarray*}$$

For more information about the Hosoya polynomial one can refer [30, 31, 32, 33, 34].

A function f(x) ϵ L²[0,1] is expanded as:

(11)

f (x) = \sum_{i = 1}^{n} c_{i} H (P_{i}, x) = C^{T} H_{P} (x),

$$\begin{equation}\label{GrindEQ__4_} f(x)=\sum _{i=1}^{n} c_{i} H(P_{i},x) =C^{T} H_P (x), \end{equation}$$

where C and H_P(x) are n × 1 matrices given by:

(12)

\begin{matrix} C = [c_{1}, c_{2}, . . ., c_{n}]^{T}, \end{matrix}

$$\begin{equation}\begin{array}{l} {C=[c_{1} ,c_{2} ,...,c_{n} ]^{T} ,} \end{array} \end{equation}$$

and

(13)

\begin{matrix} H_{P} (x) = [H (P_{1}, x), H (P_{2}, x), . . ., H (P_{n}, x)]^{T} . \end{matrix}

$$\begin{equation}\label{GrindEQ__6_} \begin{array}{l} {H_P(x)=[H(P_1,x),H(P_2,x),...,H(P_n,x)]^{T} .} \end{array} \end{equation}$$

Method of Solution

3.1

Haar Wavelet Method(HWM)

Let us consider the equation

(14)

\begin{matrix} y (x) - \int_{0}^{1} k (x, t) y (t) d t = f (x), x, t \in [0, 1] \end{matrix}

$$\begin{eqnarray}\label{4.1} y(x)-\int_{0}^{1}k(x,t)y(t)dt=f(x), \,\,\,\,\, x,t \in [0,1] \end{eqnarray}$$

and seek the solution in the form

(15)

\begin{matrix} y (x) = \sum_{i = 1}^{N} a_{i} h_{i} (x) \end{matrix}

$$\begin{eqnarray}y(x)=\sum_{i=1}^{N} a_ih_i(x)\label{4.2} \end{eqnarray}$$

where a_i’s are Haar wavelet coefficients.

Substituting (15) in (14), we get

(16)

\begin{matrix} \sum_{i = 1}^{N} a_{i} h_{i} (x) - \sum_{i = 1}^{N} a_{i} G_{i} (x) = f (x), \end{matrix}

$$\begin{eqnarray}\sum_{i=1}^{N} a_ih_i(x)- \sum_{i=1}^{N} a_iG_i(x)=f(x), \end{eqnarray}$$

where

G_{i} (x) = \int_{0}^{1} K (x, t) h_{i} (t) d t

$$G_i(x)=\int_{0}^{1} K(x,t)h_i(t)dt$$

Substituting the collocation points in Eq. (16) , we get a system of algebraic equations

(17)

\begin{matrix} \sum_{i = 1}^{N} (a_{i} h_{i} (x_{l}) - a_{i} G_{i} (x_{l})) = f (x_{l}), l = 1, 2, . . ., N \end{matrix}

$$\begin{eqnarray}\label{4.5} \sum_{i=1}^{N} \left[ a_ih_i(x_{l})- a_iG_i(x_{l})\right] =f(x_l),\,\,\,\,\, l=1,2,. . .,N \end{eqnarray}$$

Matrix form of this system of Eq. (17) is as follows,

a (H - G) = F,

$$a(H-G) = F,$$

Substituting the Coefficients a_i in Eq. (15), we get y = (y(x_l)) is the approximate solution

y = a H

$$y = a H$$

where, F = f(f(x_l)),G = (G_i(x_l))

In the present case

(18)

\begin{matrix} G_{i} (x_{l}) = \int_{0}^{1} K (x, t) h_{i} (t) d t = (\begin{matrix} x + 0.5 f o r i = 1 \\ \frac{- 1}{4 m^{2}} f o r i > 1 \end{matrix}) \end{matrix}

$$\begin{eqnarray}G_i(x_{l})=\int_{0}^{1} K(x,t) h_i(t)dt= \begin{cases} x+0.5 \,\,\,\,\,\, for \,\,\, i=1\\ \frac{-1}{4m^2} \,\,\,\,\,\, for \,\,\, i >1 \end{cases} \end{eqnarray}$$

where K(x,t) = x + t.

3.2

Hosoya Polynomial Method (HPM)

Consider the Fredholm integral equation,

(19)

y (x) = f (x) + \int_{0}^{1} K (x, t) u (t) d t, 0 \leq x t \leq 1

$$\begin{equation}\label{GrindEQ__7_} y(x)=f(x)+\int_{0}^{1}K(x,t)u(t)dt,\, \, \, \, \, 0\le x\, \, t\le 1 \end{equation}$$

To solve Eq. (19), the procedure is as follows:

Step 1: We first approximate y(x) as truncated series defined in Eq. (11) That is,

(20)

y (x) = C^{T} H_{P} (x)

$$\begin{equation}\label{GrindEQ__8_} y(x)=C^{T} H_P(x) \end{equation}$$

where C and H_P(x)are defined similarly to Eqs. (12) and (13).

Step 2: Then substituting Eq. (20) in Eq. (19), we get,

(21)

C^{T} H_{P} (x) = f (x) + \int_{0}^{1} K (x, t) [C^{T} H_{P} (t)] d t

$$\begin{equation}\label{GrindEQ__9_} C^{T} H_P(x)=f(x)+\int _{0}^{1} K(x,t) [{C^{T} H_P(t)}] dt \end{equation}$$

Step 3: Substituting the collocation point $x_{i} = \frac{i - 0.5}{n}, i = 1, 2, . . ., n$ $x_{i} =\frac{i-0.5}{n} ,\, i=1,2,...,n$in Eq. (21). Then we obtain,

(22)

C^{T} H_{P} (x_{i}) = f (x_{i}) + C^{T} \int_{0}^{1} K (x_{i}, t) H_{P} (t) d t

$$\begin{equation}\label{GrindEQ__10_} C^{T} H_P(x_i)=f(x_{i} )+C^{T}\int _{0}^{1}K(x_i,t) { H_P(t)} dt \end{equation}$$

C^{T} (H_{P} (x_{i}) - Z) = f, w h e r e Z = \int_{0}^{1} K (x_{i}, t) H_{P} (t) d t

$$C^{T} (H_P(x_i)-Z)=f,\, \, \, {\rm where}\, \, Z=\int _{0}^{1}K(x_i,t) { H_P(t)} dt$$

Step 4: Now, we get the system of algebraic equations with unknown coefficients.

C^{T} K = f, w h e r e K = (H_{P} (x_{i}) - Z)

$$C^{T} K=f,\, \, \, {\rm where}\, \, \, K=( H_P(x_i)-Z)$$

Solving the above system of equations, we get the Hosoya coefficients ‘C’ and then substituting these coefficients in Eq. (20), we get the required approximate solution of Eq. (19).

Numerical Examples

In this section, we consider the some of the demonstrate the capability of the method and error function is presented to verify the accuracy and efficiency of the following numerical results:

E r r o r f u n c t i o n = ∥ y_{e} (x_{i}) - y_{a} (x_{i}) ∥_{\infty} = \sqrt{\sum_{i = 1}^{n} (y_{e} (x_{i}) - y_{a} (x_{i}))^{2}}

$$Error function =\Vert y_{e}(x_{i})- y_{a}(x_{i})\Vert_{\infty}=\sqrt{\sum\limits_{i=1}^{n}(y_{e}(x_{i})- y_{a}(x_{i}))^2}$$

where y_e and y_a are the exact and approximate solution respectively.

Example 1.Consider Fredholm integral equation of the second kind [20],

(23)

y (x) = x^{6} l o g (x) + \int_{0}^{1} (x + t) y (t) d t, 0 \leq x \leq 1.

$$\begin{equation}\label{GrindEQ__11_} y(x)=x^{6}log(x)+ \int_{0}^{1}(x+t)y(t)dt, \, \, \, \, 0\le x\le 1. \end{equation}$$

which has the exact solution y(x) = x⁶log(x)+0.3096x+0.1752.

Table 1

Numerical result of Example 1.

x	Exact solution	HWM (N = 8)	Abs. Error (HWM)	HPM (n = 8)	Abs. Error (HPM)
0.0625	0.1945	0.2046	1.00e-02	0.1947	1.94e-04
0.1875	0.2332	0.2453	1.20e-02	0.2334	2.36e-04
0.3125	0.2709	0.2850	1.41e-02	0.2711	2.78e-04
0.4375	0.3049	0.3211	1.62e-02	0.3052	3.19e-04
0.5625	0.3311	0.3495	1.83e-02	0.3315	3.61e-04
0.6875	0.3485	0.3689	2.04e-02	0.3489	4.03e-04
0.8125	0.3670	0.3895	2.25e-02	0.3675	4.45e-04
0.9375	0.4216	0.4462	2.45e-02	0.4221	4.86e-04

Example 2. Consider the Fredholm integral equation of the second kind [2],

(24)

y (x) = x^{2} + \int_{0}^{1} (x + t) y (t) d t, 0 \leq x \leq 1.

$$\begin{equation} y(x)=x^{2}+ \int_{0}^{1}(x+t)y(t)dt, \, \, \, \, 0\le x\le 1. \end{equation}$$

which has the exact solution y(x) = x²-5x-(17/6). We solved the Eq. (24) by Hosoya polynomial method for n = 3, we get the Hosoya coefficient $C_{1} = \frac{49}{6}, C_{2} = - 7, C_{3} = 1$ $C_1=\dfrac{49}{6}, C_2=-7,C_3=1$. Substituting these coefficients in Eq. (20) we obtain,

\begin{matrix} y (x) = \frac{49}{6} & H_{P} (x_{1}) + (- 7) H_{P} (x_{2}) + (1) H_{P} (x_{3}) \\ y (x) = x^{2} - 5 x - (17 / 6) \end{matrix}

$$\eqalign{y(x)=\dfrac{49}{6}& H_P(x_1)+(-7) H_P(x_2)+(1)H_P(x_3)\\ &y(x)=x^2-5x-(17/6)} $$

Solved by analyticaly, which gives the exact solution of this problem.

Example 3. Consider the Fredholm integral equation of the second kind [3],

(25)

y (x) = e x p (x) + (1 - e x p (1)) x - 1 + \int_{0}^{1} (x + t) y (t) d t, 0 \leq x \leq 1.

$$\begin{equation}\label{GrindEQ__28_} y(x)=exp(x)+(1-exp(1))x-1+ \int_{0}^{1}(x+t)y(t)dt, \, \, \, \, 0\le x\le 1. \end{equation}$$

which has the exact solution y(x) = exp(x).

Table 2

Numerical result of Example 3.

x	Exact solution	HWM (N = 8)	Abs. Error (HWM)	HPM (n = 8)	Abs. Error (HPM)
0.0625	1.0645	1.0892	2.47e-02	1.0645	5.50e-09
0.1875	1.2062	1.2362	2.99e-02	1.2062	6.56e-09
0.3125	1.3668	1.4021	3.52e-02	1.3668	7.63e-09
0.4375	1.5488	1.5893	4.04e-02	1.5488	8.69e-09
0.5625	1.7551	1.8008	4.57e-02	1.7551	9.77e-09
0.6875	1.9887	2.0397	5.09e-02	1.9887	1.08e-08
0.8125	2.2535	2.3098	5.62e-02	2.2535	1.19e-08
0.9375	2.5536	2.6151	6.14e-02	2.5536	1.30e-08

Table 3

Error Analysis.

Haar Wavelet Method (HWM)				Hosoya Polynomial Method (HPM)
N	Example 1	Example 2	Example 3	n	Example 1	Example 3
8	2.46e-02	7.2e-02	2.17e-02	3	1.25e-01	1.83e-02
16	6.70e-03	1.7e-02	1.55e-02	4	4 6.89e-02	2.15e-03
32	2.10e-03	4.3e-03	3.90e-03	6	6 5.64e-03	6.86e-06
64	9.25e-04	1.3e-02	9.82e-04	8	8 4.86e-04	1.30e-08

Example 4. Consider the Fredholm integral equation of the second kind [19],

(26)

y (x) = e^{2 x + (\frac{1}{3})} + \int_{0}^{1} \frac{- 1}{3} e^{2 x - (\frac{5}{3}) t} y (t) d t, 0 \leq x \leq 1.

$$\begin{equation} y(x)=e^{2x+(\frac{1}{3})}+ \int_{0}^{1}\frac{-1}{3}e^{2x-(\frac{5}{3})t}y(t)dt, \, \, \, \, 0\le x\le 1. \end{equation}$$

which has the exact solution $y (x) = e^{2 x} .$ $y(x)=e^{2x}.$Applying the proposed method to solve Eq. (26) for n = 8. We obtain the approximate solution y(x) as shown in table 4. Error analysis is shown in figure 1.

Table 4

Numerical result of Example 4.

x	Exact solution	Method [19] (m = 128	Abs. Error (Method [19])	HPM (n = 8)	Abs. Error (HPM)
0	1.00000	1.00004	4.00e-05	0.99999	7.39e-06
0.1	1.22140	1.22147	6.72e-05	1.22140	6.61e-07
0.2	1.49182	1.49180	2.46e-05	1.49182	1.35e-09
0.3	1.82211	1.82218	6.12e-05	1.82211	4.61e-08
0.4	2.22554	2.22551	3.09e-05	2.22554	1.29e-07
0.5	2.71828	2.71822	6.18e-05	2.71828	6.17e-08
0.6	3.32011	3.32017	5.30e-05	3.32011	1.73e-07
0.7	4.05519	4.05529	9.00e-05	4.05519	1.31e-07
0.8	4.95303	4.95306	2.75e-05	4.95303	1.25e-07
0.9	6.04964	6.04961	3.74e-05	6.04964	9.70e-07
1	7.38905	7.38901	4.60e-05	7.38905	8.99e-06

Example 5. Consider the Fredholm integral equation of the second kind [4],

(27)

y (x) = e^{x} - \frac{e^{x + 1} - 1}{x + 1} + \int_{0}^{1} e^{x t} y (t) d t, 0 \leq x \leq 1.

$$\begin{equation} y(x)=e^x-\frac{e^{x+1}-1}{x+1}+ \int_{0}^{1}e^{xt} y(t)dt, \, \, \, \, 0\le x\le 1. \end{equation}$$

which has the exact solution y(x) = e^x. Applying the proposed method to solve Eq. (27) for N = 8. We obtain the approximate solution y(x) as shown in table 5. Error analysis is shown in figure 2.

Table 5

Numerical result of Example 5.

x	Exact solution	Method [4] (k = 32)	Abs. Error (Method [4])	HPM (n = 8)	Abs. Error (HPM)
0	1.00000	1.01642	1.64e-02	1.00000	3.79e-06
0.1	1.10517	1.11627	1.10e-02	1.10517	1.82e-07
0.2	1.22140	1.22593	4.52e-03	1.22140	4.30e-07
0.3	1.34985	1.34637	3.48e-03	1.34985	4.56e-07
0.4	1.49182	1.47864	1.31e-03	1.49182	4.88e-07
0.5	1.64872	1.62391	2.48e-02	1.64872	5.39e-07
0.6	1.82211	1.84004	1.79e-02	1.82211	5.80e-07
0.7	2.01375	2.02082	7.06e-03	2.01375	6.34e-07
0.8	2.22554	2.21936	6.18e-03	2.22554	6.90e-07
0.9	2.45960	2.43742	2.21e-02	2.45960	7.18e-07
1	2.71828	2.67690	4.13e-02	2.71828	1.10e-06

Conclusion

A newly developed method Hosoya polynomial method (HPM) [35] is applied for the numerical solution of Fredholm integral equations and compared with the Haar Wavelet Method (HWM)). Both the methods reduces an integral equation into a set of algebraic equations. Numerical results of the Hosoya polynomial method (HPM) gives higher accuracy with exact ones and existing method (HWM). Some illustrative examples are tested through the error analysis. This shows that, the accuracy improves with increasing the n number of vertices of polynomial for better accuracy. Error analysis justifies the comparative study of a new developed method is effective, validity and applicability with existing method.

eISSN:: 2444-8656
Language:: English

Publication timeframe:: Volume Open
Journal Subjects:: Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics

Journal RSS Feed

A Comparative Study on Haar Wavelet and Hosaya Polynomial for the numerical solution of Fredholm integral equations

Published Online: Sep 05, 2018

Page range: 447 - 458

Received: Apr 28, 2018

Accepted: Sep 05, 2018

DOI: https://doi.org/10.21042/AMNS.2018.2.00035

Keywords
Fredholm integral equation, Hosoya polynomial, Haar Wavelet, path

© 2018 S.C. Shiralashetti et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Fig. 1

Fig. 2

A Comparative Study on Haar Wavelet and Hosaya Polynomial for the numerical solution of Fredholm integral equations

Published Online: Sep 05, 2018

Page range: 447 - 458

Received: Apr 28, 2018

Accepted: Sep 05, 2018

DOI: https://doi.org/10.21042/AMNS.2018.2.00035

KeywordsFredholm integral equation, Hosoya polynomial, Haar Wavelet, path

© 2018 S.C. Shiralashetti et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Fig. 1

Fig. 2

Keywords
Fredholm integral equation, Hosoya polynomial, Haar Wavelet, path