Polynomial interpolation
In numerical analysis, polynomial interpolation is the interpolation of a given data set by the polynomial of lowest possible degree that passes through the points in the dataset.
≈ 24 min read
Names in other languages
No names in other languages are held for this entry yet.
Article bodies are English and are not machine-translated. A name in another language is not a translation of the article.
In numerical analysis, polynomial interpolation is the interpolation of a given data set by the polynomial of lowest possible degree that passes through the points in the dataset. Given a set of data points (x_0,y_0), \ldots, (x_n,y_n) , with no two x_j the same, a polynomial function p(x)=a_0+a_1x+\cdots+a_nx^n is said to interpolate the data if p(x_j)=y_j for each j\in\{0,1,\dotsc,n\} .
There is always a unique such polynomial, commonly given by two explicit formulas, the Lagrange polynomials and Newton polynomials. In numerical analysis, polynomial interpolation is essential to perform sub-quadratic multiplication and squaring, such as Karatsuba multiplication and Toom–Cook multiplication, where interpolation through points on a product polynomial yields the specific product required.
For example, given a = f(x) = a 0 x 0 + a 1 x 1 + ··· and b = g(x) = b 0 x 0 + b 1 x 1 + ···, the product ab is a specific value of W(x) = f(x)g(x). One may easily find points along W(x) at small values of x, and interpolation based on those points will yield the terms of W(x) and the specific product ab. As fomulated in Karatsuba multiplication, this technique is substantially faster than quadratic multiplication, even for modest-sized inputs, especially on parallel hardware.
In computer science, polynomial interpolation also leads to algorithms for secure multi party computation and secret sharing.
Interpolation theorem
For any n+1 bivariate data points (x_0,y_0),\dotsc,(x_n,y_n) \in \R^2 , where no two x_j are the same, there exists a unique polynomial p(x) of degree at most n that interpolates these points, i.e. p(x_0)=y_0, \ldots, p(x_n)=y_n . Equivalently, for a fixed choice of interpolation nodes x_j , polynomial interpolation defines a linear bijection L_n between the (n+1)-tuples of real-number values (y_0,\ldots,y_n)\in \R^{n+1} and the vector space P(n) of real polynomials of degree at most n: L_n : \mathbb{R}^{n+1} \stackrel{\sim}{\longrightarrow}\, P(n). This is a type of unisolvence theorem. The theorem is also valid over any infinite field in place of the real numbers \R , for example the rational or complex numbers.
First proof
Consider the Lagrange basis functions L_0(x),\ldots,L_n(x) given by: L_j(x)=\prod_{i=0,i\neq j}^{n}\frac{x-x_i}{x_j-x_i} = \frac{(x-x_0)\cdots(x-x_{j-1})(x-x_{j+1})\cdots(x-x_n)} {(x_j-x_0)\cdots(x_j-x_{j-1})(x_j-x_{j+1})\cdots(x_j-x_n)}. Notice that L_j(x) is a polynomial of degree n , and we have L_j(x_k)=0 for each j\neq k , while L_k(x_k)=1 .
It follows that the linear combination: p(x) = \sum_{j=0}^n y_j L_j(x) has p(x_k)=\sum_j y_j \,L_j(x_k) = y_k , so p(x) is an interpolating polynomial of degree n . To prove uniqueness, assume that there exists another interpolating polynomial q(x) of degree at most n , so that p(x_k)=q(x_k) for all k=0,\dotsc,n . Then p(x)-q(x) is a polynomial of degree at most n which has n+1 distinct zeros (the x_k ). But a non-zero polynomial of degree at most n can have at most n zeros, so p(x)-q(x) must be the zero polynomial, i.e. p(x)=q(x) .
Second proof
Write out the interpolation polynomial in the form Substituting this into the interpolation equations p(x_j) = y_j , we get a system of linear equations in the coefficients a_j , which reads in matrix-vector form as the following multiplication: \begin{bmatrix} x_0^n & x_0^{n-1} & x_0^{n-2} & \ldots & x_0 & 1 \\ x_1^n & x_1^{n-1} & x_1^{n-2} & \ldots & x_1 & 1 \\ \vdots & \vdots & \vdots & & \vdots & \vdots \\ x_n^n & x_n^{n-1} & x_n^{n-2} & \ldots & x_n & 1 \end{bmatrix} \begin{bmatrix} a_n \\ a_{n-1} \\ \vdots \\ a_0 \end{bmatrix} = \begin{bmatrix} y_0 \\ y_1 \\ \vdots \\ y_n \end{bmatrix}. An interpolant p(x) corresponds to a solution A = (a_n,\ldots,a_0) of the above matrix equation X \cdot A = Y . The matrix X on the left is a Vandermonde matrix, whose determinant is known to be \textstyle \det(X) = \prod_{0 \le i which is non-zero since the nodes x_j are all distinct.
This ensures that the matrix is invertible and the equation has the unique solution A = X^{-1}\cdot Y ; that is, p(x) exists and is unique.
Corollary
If f(x) is a polynomial of degree at most n , then the interpolating polynomial of f(x) at n+1 distinct points is f(x) itself.
Constructing the interpolation polynomial
\begin{align} p(x) &= \frac{(x-x_1)(x-x_2)\cdots(x-x_n)}{(x_0-x_1)(x_0-x_2)\cdots(x_0-x_n)} y_0 \\ [4pt] &+ \frac{(x-x_0)(x-x_2)\cdots(x-x_n)}{(x_1-x_0)(x_1-x_2) \cdots(x_1-x_n)}y_1 \\ [4pt] &+ \cdots\\ [4pt] &+\frac{(x-x_0)(x-x_1)\cdots(x-x_{n-1})}{(x_n-x_0)(x_n-x_1)\cdots(x_n-x_{n-1})}y_n \\ [7pt] &=\sum_{i=0}^n \Biggl( \prod_{\stackrel{\!0\,\leq\, j\,\leq\, n}{j\,\neq\, i}} \frac{x-x_j}{x_i-x_j} \Biggr) y_i =\sum_{i=0}^n \frac{p(x)}{p'(x_i)(x-x_i)}\,y_i \end{align} For matrix arguments, this formula is called Sylvester's formula and the matrix-valued Lagrange polynomials are the Frobenius covariants. For a polynomial p_n of degree less than or equal to n , that interpolates f at the nodes x_i where i = 0,1,2,3,\cdots,n . Let p_{n+1} be the polynomial of degree less than or equal to n+1 that interpolates f at the nodes x_i where i = 0,1,2,3,\cdots,n, n+1 .
Then p_{n+1} is given by: p_{n+1}(x) = p_n(x) +a_{n+1}w_n(x) where w_n(x) := \prod_{i=0}^n (x-x_i) also known as Newton basis and a_{n+1} :={f(x_{n+1})-p_n(x_{n+1}) \over w_n(x_{n+1})} . Proof: This can be shown for the case where i = 0,1,2,3,\cdots,n : p_{n+1}(x_i) = p_n(x_i) +a_{n+1}\prod_{j=0}^n (x_i-x_j) = p_n(x_i) and when i = n+1 : p_{n+1}(x_{n+1}) = p_n(x_{n+1}) +{f(x_{n+1})-p_n(x_{n+1}) \over w_n(x_{n+1})} w_n(x_{n+1}) = f(x_{n+1}) By the uniqueness of interpolated polynomials of degree less than n+1 , p_{n+1}(x) = p_n(x) +a_{n+1}w_n(x) is the required polynomial interpolation.
The function can thus be expressed as: p_{n}(x) = a_0+a_1(x-x_0)+a_2(x-x_0)(x-x_1)+\cdots + a_n(x-x_0)\cdots(x-x_{n-1}) . To find a_i , we have to solve the lower triangular matrix formed by arranging p_{n} (x_i)=f(x_i)=y_i from above equation in matrix form: \begin{bmatrix} 1 & & \ldots & & 0 \\ 1 & x_1-x_0 & & & \\ 1 & x_2-x_0 & (x_2-x_0)(x_2-x_1) & & \vdots \\ \vdots & \vdots & & \ddots & \\ 1 & x_n-x_0 & \ldots & \ldots & \prod_{j=0}^{n-1}(x_n - x_j) \end{bmatrix} \begin{bmatrix} a_0 \\ \\ \vdots \\ \\ \\ a_{n} \end{bmatrix} = \begin{bmatrix} y_0 \\ \\ \vdots \\ \\ \\ y_{n} \end{bmatrix} The coefficients are derived as a_j := [y_0,\ldots,y_j] where [y_0,\ldots,y_j] is the notation for divided differences.
Thus, Newton polynomials are used to provide a polynomial interpolation formula of n points. If x_0, x_1, \dots, x_k are consecutively arranged and equally spaced with {x}_{i}={x}_{0}+ih for i = 0, 1, ..., k and some variable x is expressed as {x}={x}_{0}+sh , then the difference x-x_i can be written as (s-i)h . So the Newton polynomial becomes \begin{align} N(x) &= [y_0] + [y_0,y_1]sh + \cdots + [y_0,\ldots,y_k] s (s-1) \cdots (s-k+1){h}^{k} \\ &= \sum_{i=0}^{k}s(s-1) \cdots (s-i+1){h}^{i}[y_0,\ldots,y_i] \\ &= \sum_{i=0}^{k}{s \choose i}i!{h}^{i}[y_0,\ldots,y_i]. \end{align} Since the relationship between divided differences and forward differences is given as: [y_j, y_{j+1}, \ldots , y_{j+n}] = \frac{1}{n!h^n}\Delta^{(n)}y_j, Taking y_i=f(x_i) , if the representation of x in the previous sections was instead taken to be x=x_j+sh , the Newton forward interpolation formula is expressed as: f(x) \approx N(x)=N(x_j+sh) = \sum_{i=0}^{k}{s \choose i}\Delta^{(i)} f(x_j) which is the interpolation of all points after x_j . It is expanded as: f(x_j+sh)=f(x_j)+\frac{s}{1!}\Delta f(x_j)+ \frac{s(s-1)}{2!}\Delta^2 f(x_j)+\frac{s(s-1)(s-2)}{3!}\Delta^3 f(x_j)+\frac{s(s-1)(s-2)(s-3)}{4!}\Delta^4 f(x_j)+\cdots N(x)=[y_k]+[{y}_{k}, {y}_{k-1}](x-{x}_{k})+\cdots+[{y}_{k},\ldots,{y}_{0}](x-{x}_{k})(x-{x}_{k-1})\cdots(x-{x}_{1}).
If {x}_{k},\;{x}_{k-1},\;\dots,\;{x}_{0} are equally spaced with {x}_{i}={x}_{k}-(k-i)h for i = 0, 1, ..., k and {x}={x}_{k}+sh , then, \begin{align} N(x) &= [{y}_{k}]+ [{y}_{k}, {y}_{k-1}]sh+\cdots+[{y}_{k},\ldots,{y}_{0}]s(s+1)\cdots(s+k-1){h}^{k} \\ &=\sum_{i=0}^{k}{(-1)}^{i}{-s \choose i}i!{h}^{i}[{y}_{k},\ldots,{y}_{k-i}]. \end{align} Since the relationship between divided differences and backward differences is given as: [{y}_{j}, y_{j-1},\ldots,{y}_{j-n}] = \frac{1}{n!h^n}\nabla^{(n)}y_j, taking y_i=f(x_i) , if the representation of x in the previous sections was instead taken to be x=x_j+sh , the Newton backward interpolation formula is expressed as: f(x) \approx N(x) =N(x_j+sh)=\sum_{i=0}^{k}{(-1)}^{i}{-s \choose i}\nabla^{(i)} f(x_j). which is the interpolation of all points before x_j . It is expanded as: f(x_j+sh)=f(x_j)+\frac{s}{1!}\nabla f(x_j)+ \frac{s(s+1)}{2!}\nabla^2 f(x_j)+\frac{s(s+1)(s+2)}{3!}\nabla^3 f(x_j)+\frac{s(s+1)(s+2)(s+3)}{4!}\nabla^4 f(x_j)+\cdots The factors are expressed using the formula: C(u+k,n)=\frac{(u+k)(u+k-1)\cdots(u+k-n+1)}{n!} If a path goes from \Delta^{n-1}y_s to \Delta^{n+1}y_{s-1} , it can connect through three intermediate steps, (a) through \Delta^{n}y_{s-1} , (b) through C(u-s ,n) or (c) through \Delta^{n}y_s . Proving the equivalence of these three two-step paths should prove that all (n-step) paths can be morphed with the same starting and ending, all of which represents the same formula. Path (a): C(u-s, n) \Delta^n y_{s-1}+C(u-s+1, n+1) \Delta^{n+1} y_{s-1} Path (b): C(u-s, n) \Delta^n y_s + C(u-s, n+1) \Delta^{n+1} y_{s-1} Path (c): C(u-s, n) \frac{\Delta^n y_{s-1}+\Delta^n y_{s}}{2} \quad+\frac{C(u-s+1, n+1)+C(u-s, n+1)}{2} \Delta^{n+1} y_{s-1} Subtracting contributions from path a and b: \begin{aligned} \text{Path a - Path b}= & C(u-s, n)(\Delta^n y_{s-1}-\Delta^n y_s) +(C(u-s+1, n+1)-C(u-s, n-1)) \Delta^{n+1} y_{s-1} \\ = & - C(u-s, n)\Delta^{n+1} y_{s-1} + C(u-s, n) \frac{(u-s+1)-(u-s-n)}{n+1} \Delta^{n+1} y_{s-1} \\ = & C(u-s, n)(-\Delta^{n+1} y_{s-1}+\Delta^{n+1} y_{s-1} )=0 \\ \end{aligned} Thus, the contribution of either path (a) or path (b) is the same.
Since path (c) is the average of path (a) and (b), it also contributes identical function to the polynomial. Hence the equivalence of paths with same starting and ending points is shown.
To check if the paths can be shifted to different values in the leftmost corner, taking only two step paths is sufficient: (a) y_{s+1} to y_{s} through \Delta y_{s} or (b) factor between y_{s+1} and y_{s} , to y_{s} through \Delta y_{s} or (c) starting from y_{s} . Path (a) y_{s+1}+C(u-s-1,1) \Delta y_s - C(u-s, 1) \Delta y_s Path (b) \frac{y_{s+1}+y_s}{2}+\frac{C(u-s-1,1)+C(u-s, 1)}{2} \Delta y_s - C(u-s, 1) \Delta y_s Path (c) y_{s} Since \Delta y_{s} = y_{s+1}-y_s , substituting in the above equations shows that all the above terms reduce to y_{s} and are hence equivalent.
Hence these paths can be morphed to start from the leftmost corner and end in a common point. y(s) &=y_0+C(s, 1) \Delta y_0+C(s, 2) \Delta^2 y_0+C(s, 3) \Delta^3 y_0+\cdots \\ & =y_0+s \Delta y_0+\frac{s(s-1)}{2} \Delta^2 y_0+\frac{s(s-1)(s-2)}{3 !} \Delta^3 y_0+\frac{s(s-1)(s-2)(s-3)}{4 !} \Delta^4 y_0+\cdots \end{aligned} whereas, taking positive slope transversal from y_n to \nabla^n y_n = \Delta^n y_0 , gives the interpolation formula of all the n+1 consecutively arranged points, equivalent to Newton's backward interpolation formula: \begin{aligned} y(u) & = y_k+C(u-k, 1) \Delta y_{k-1}+C(u-k+1,2) \Delta^2 y_{k-2} +C(u - k+2,3) \Delta^3 y_{k-3}+\cdots \\ & = y_k+(u-k) \Delta y_{k-1} +\frac{(u-k+1) (u-k)}{2} \Delta^2 y_{k-2}+\frac{(u-k+2)(u-k+1)(u-k)}{3 !} \Delta^3 y_{k-3}+\cdots \\ y(k+s) & = y_k+(s) \nabla y_{k} +\frac{(s+1) s}{2} \nabla^2 y_{k}+\frac{(s+2)(s+1) s}{3 !} \nabla^3 y_{k}+\frac{(s+3)(s+2)(s+1) s}{4 !} \nabla^4 y_{k}+\cdots \\ \end{aligned} where s=u-k is the number corresponding to that introduced in Newton interpolation. y(u)=y_0+u \Delta y_0+\frac{u(u-1)}{2} \Delta^2 y_{-1} +\frac{(u+1)u\left(u-1\right)}{3 !} \Delta^3 y_{-1}+ \frac{(u+1)u\left(u-1\right)(u-2)}{4 !} \Delta^4 y_{-2} + \cdots whereas starting from y_0 with positive slope, we get Gauss backward formula: y(u)=y_0+u \Delta y_{-1}+\frac{(u+1)u}{2} \Delta^2 y_{-1} +\frac{(u+1)u\left(u-1\right)}{3 !} \Delta^3 y_{-2}+ \frac{(u+2)(u+1)u\left(u-1\right)}{4 !} \Delta^4 y_{-2} + \cdots y(u)&= y_0 +u \frac{\Delta y_0+\Delta y_{-1}}{2}+\frac{C(u+1,2)+C(u, 2)}{2} \Delta^2 y_{-1} +C(u+1,3) \frac{\Delta^3 y_{-2}+\Delta^3 y_{-1}}{2}+\cdots \\ & = y_0+u \frac{\Delta y_0+\Delta y_{-1}}{2}+\frac{u^2}{2} \Delta^2 y_{-1}+\frac{u(u^2-1)}{3 !} \frac{\Delta^3 y_{-2}+\Delta^3 y_{-1}}{2}+\frac{u^2(u^2-1)}{4!}\Delta^4 y_{-2}+\cdots \end{aligned} Stirling formula is the average of Gauss forward and Gauss backward formulas. y(u)&=1{\frac{y_{0}+y_{1}}{2}}+{\frac{C(u,1)+C(u-1,1)}{2}}\Delta y_{0}+C(u,2){\frac{\Delta^{2}y_{-1}+\Delta^{2}y_{0}}{2}}+\cdots\\ &= \frac{y_{0}+y_{1}}{2}+\left(u-{\frac{1}{2}}\right)\Delta y_{0}+\frac{u(u-1)}{2}\frac{\Delta^{2}y_{-1} +\Delta^{2}y_{0}}{2}+\frac{\left(u-{\frac{1}{2}}\right)u\left( u-1\right)}{3!}\Delta^{3} y_{0} + \frac{(u+1)u(u-1)(u-2)}{4!}\frac{\Delta^{4}y_{-1}+\Delta^{4}y_{-2}}{2}+\cdots\\ \end{align} Another method is preferred when the aim is not to compute the coefficients of p(x), but only a single value p(a) at a point x = a not in the original data set. The Lagrange form computes the value p(a) with complexity O(n 2 ).
The Bernstein form was used in a constructive proof of the Weierstrass approximation theorem by Bernstein and has gained great importance in computer graphics in the form of Bézier curves. y(x) := \sum_{j=0}^{k} y_j c_j(x) with each coefficient c_j(x) given by the corresponding Lagrange basis polynomial on the given positions x_j : c_j(x) = L_j(x_0,\ldots,x_n;x) = \prod_{ 0 \le i \le n \atop i \neq j } \frac{x-x_i}{x_j-x_i} = \frac{(x-x_0)}{(x_j-x_0)} \cdots \frac{(x-x_{j-1})}{(x_j-x_{j-1})} \frac{(x-x_{j+1})}{(x_j-x_{j+1})} \cdots \frac{(x-x_n)}{(x_j-x_n)}. Since the coefficients depend only on the positions x_j , not the values y_j , we can use the same coefficients to find the interpolating polynomial for a second set of data points (x_0, v_0), \ldots, (x_n, v_n) at the same positions: v(x) := \sum_{j=0}^{k} v_j c_j(x). Furthermore, the coefficients c_j(x) only depend on the relative spaces x_i-x_j between the positions.
Thus, given a third set of data whose points are given by the new variable t = ax+b (an affine transformation of x , inverted by x=\tfrac{t-b}{a} ): (t_0, w_0), \ldots, (t_j, w_j) \ldots, (t_n, w_n) \qquad \text{with}\qquad t_j = ax_j + b, we can use a transformed version of the previous coefficient polynomials: \tilde c_j(t) := c_j(\tfrac{t-b}{a}) = c_j(x), and write the interpolation polynomial as: w(t) := \sum_{j=0}^{k} w_j \tilde c_j(t). Data points (x_j,y_j) often have equally spaced positions, which may be normalized by an affine transformation to x_j = j . For example, consider the data points (0,y_0), (1,y_1), (2,y_2) . The interpolation polynomial in the Lagrange form is the linear combination \begin{align} y(x) := \sum_{j=0}^2 y_j c_j(x) &= y_0 \frac{(x-1)(x-2)}{(0-1)(0-2)} + y_1 \frac{(x-0)(x-2)}{(1-0)(1-2)} + y_2 \frac{(x-0)(x-1)}{(2-0)(2-1)} \\ &= \tfrac12 y_0 (x-1)(x-2) - y_1 (x-0)(x-2) + \tfrac12 y_2 (x-0)(x-1). \end{align} For example, y(3) = y_3 = y_0 - 3y_1 + 3y_2 and y(1.5) = y_{1.5} = \tfrac18 (-y_0 + 6y_1 + 3y_2) .
The case of equally spaced points can also be treated by the method of finite differences. The first difference of a sequence of values v=\{v_j\}_{j=0}^\infty is the sequence \Delta v = u = \{u_j\}_{j=0}^\infty defined by u_j = v_{j+1}-v_j . Iterating this operation gives the n th difference operation \Delta^n v = u , defined explicitly by u_j = \sum_{k=0}^n (-1)^{n-k} {n\choose k} v_{j+k}.
A polynomial y(x) of degree d defines a sequence of values at positive integer points, y_j = y(j) , and the (d+1)^{\text{th}} difference of this sequence is identically zero: \Delta^{d+1} y = 0 . Thus, given values y_0,\ldots,y_n at equally spaced points, where n=d+1 , we have: (-1)^n y_0 + (-1)^{n-1} \binom{n}{1} y_1 + \cdots - \binom n{n-1} y_{n-1} + y_n= 0.
For example, 4 equally spaced data points y_0,y_1,y_2,y_3 of a quadratic y(x) obey 0 = -y_0 + 3y_1 - 3y_2 + y_3 , and solving for y_3 gives the same interpolation equation obtained above using the Lagrange method.
Interpolation error: Lagrange remainder formula
When interpolating a given function f by a polynomial p_n of degree at the nodes x 0 ,..., x n we get the error f(x) - p_n(x) = f[x_0,\ldots,x_n,x] \prod_{i=0}^n (x-x_i) where f[x_0,\ldots,x_n,x] is the (n+1) st divided difference of the data points (x_0,f(x_0)),\ldots,(x_n,f(x_n)),(x,f(x)) . Furthermore, there is a Lagrange remainder form of the error, for a function f which is times continuously differentiable on a closed interval I , and a polynomial p_n(x) of degree at most that interpolates f at distinct points x_0,\ldots,x_n\in I . For each x\in I there exists \xi\in I such that f(x) - p_n(x) = \frac{f^{(n+1)}(\xi)}{(n+1)!} \prod_{i=0}^n (x-x_i). This error bound suggests choosing the interpolation points to minimize the product \left | \prod (x - x_i) \right | , which is achieved by the Chebyshev nodes. Set the error term as R_n(x) = f(x) - p_n(x) , and define an auxiliary function: Y(t) = R_n(t) - \frac{R_n(x)}{W(x)} W(t) \qquad\text{where}\qquad W(t) = \prod_{i=0}^n (t-x_i). Thus: Y^{(n+1)}(t) = R_n^{(n+1)}(t) - \frac{R_n(x)}{W(x)} \ (n+1)!
But since p_n(x) is a polynomial of degree at most , we have R_n^{(n+1)}(t) = f^{(n+1)}(t) , and: Y^{(n+1)}(t) = f^{(n+1)}(t) - \frac{R_n(x)}{W(x)} \ (n+1)! Now, since are roots of R_n(t) and W(t) , we have Y(x)=Y(x_j)=0 , which means has at least roots. From Rolle's theorem, Y^\prime(t) has at least roots, and iteratively Y^{(n+1)}(t) has at least one root in the interval .
Thus: Y^{(n+1)}(\xi) = f^{(n+1)}(\xi) - \frac{R_n(x)}{W(x)} \ (n+1)! = 0 and: R_n(x) = f(x) - p_n(x) = \frac{f^{(n+1)}(\xi)}{(n+1)!} \prod_{i=0}^n (x-x_i) . This parallels the reasoning behind the Lagrange remainder term in the Taylor theorem; in fact, the Taylor remainder is a special case of interpolation error when all interpolation nodes are identical. Note that the error will be zero when x = x_i for any i. Thus, the maximum error will occur at some point in the interval between two successive nodes.
Equally spaced intervals
In the case of equally spaced interpolation nodes where x_i = a + ih , for i=0,1,\ldots,n, and where h = (b-a)/n, the product term in the interpolation error formula can be bound as \left|\prod_{i=0}^n (x-x_i)\right| = \prod_{i=0}^n \left|x-x_i\right| \leq \frac{n!}{4} h^{n+1}. Thus the error bound can be given as \left|R_n(x)\right| \leq \frac{h^{n+1}}{4(n+1)} \max_{\xi\in[a,b]} \left|f^{(n+1)}(\xi)\right| However, this assumes that f^{(n+1)}(\xi) is dominated by h^{n+1} , i.e. f^{(n+1)}(\xi) h^{n+1} \ll 1 .
In several cases, this is not true and the error actually increases as (see Runge's phenomenon). That question is treated in the section Convergence properties.
Lebesgue constants
We fix the interpolation nodes x 0 , ..., x n and an interval [a, b] containing all the interpolation nodes. The process of interpolation maps the function f to a polynomial p. This defines a mapping X from the space C([a, b]) of all continuous functions on [a, b] to itself. The map X is linear and it is a projection on the subspace P(n) of polynomials of degree n or less.
The Lebesgue constant L is defined as the operator norm of X. One has (a special case of Lebesgue's lemma): \left\|f-X(f)\right\| \le (L+1) \left\|f-p^*\right\|. In other words, the interpolation polynomial is at most a factor (L + 1) worse than the best possible approximation. This suggests that we look for a set of interpolation nodes that makes L small.
In particular, we have for Chebyshev nodes: L \le \frac2\pi \log(n+1) + 1. We conclude again that Chebyshev nodes are a very good choice for polynomial interpolation, as the growth in n is exponential for equidistant nodes. However, those nodes are not optimal.
Convergence properties
It is natural to ask, for which classes of functions and for which interpolation nodes the sequence of interpolating polynomials converges to the interpolated function as ? Convergence may be understood in different ways, e.g. pointwise, uniform or in some integral norm.
The situation is rather bad for equidistant nodes, in that uniform convergence is not even guaranteed for infinitely differentiable functions. One classical example, due to Carl Runge, is the function f(x) = 1 / (1 + x 2 ) on the interval . The interpolation error grows without bound as .
Another example is the function f(x) = |x| on the interval , for which the interpolating polynomials do not even converge pointwise except at the three points x = ±1, 0. One might think that better convergence properties may be obtained by choosing different interpolation nodes.
The following result seems to give a rather encouraging answer: The defect of this method, however, is that interpolation nodes should be calculated anew for each new function f(x), but the algorithm is hard to be implemented numerically. Does there exist a single table of nodes for which the sequence of interpolating polynomials converge to any continuous function f(x)? The answer is unfortunately negative: The proof essentially uses the lower bound estimation of the Lebesgue constant, which we defined above to be the operator norm of X n (where X n is the projection operator on Π n ). Now we seek a table of nodes for which \lim_{n \to \infty} X_n f = f,\text{ for every }f \in C([a,b]). Due to the Banach–Steinhaus theorem, this is only possible when norms of X n are uniformly bounded, which cannot be true since we know that \|X_n\|\geq \tfrac{2}{\pi} \log(n+1)+C. For example, if equidistant points are chosen as interpolation nodes, the function from Runge's phenomenon demonstrates divergence of such interpolation.
Note that this function is not only continuous but even infinitely differentiable on . For better Chebyshev nodes, however, such an example is much harder to find due to the following result:
Related concepts
Runge's phenomenon shows that for high values of , the interpolation polynomial may oscillate wildly between the data points. This problem is commonly resolved by the use of spline interpolation. Here, the interpolant is not a polynomial but a spline: a chain of several polynomials of a lower degree.
Interpolation of periodic functions by harmonic functions is accomplished by Fourier transform. This can be seen as a form of polynomial interpolation with harmonic base functions, see trigonometric interpolation and trigonometric polynomial.
Hermite interpolation problems are those where not only the values of the polynomial p at the nodes are given, but also all derivatives up to a given order. This turns out to be equivalent to a system of simultaneous polynomial congruences, and may be solved by means of the Chinese remainder theorem for polynomials.
Birkhoff interpolation is a further generalization where only derivatives of some orders are prescribed, not necessarily all orders from 0 to a k. Collocation methods for the solution of differential and integral equations are based on polynomial interpolation. The technique of rational function modeling is a generalization that considers ratios of polynomial functions.
At last, multivariate interpolation for higher dimensions.
See also
Newton series Polynomial regression Spline smoothing
Further reading
J. L. Walsh: Interpolation and Approximation by Rational Functions in the Complex Domain, AMS (Colloquium Publications, Vol.20), ISBN 0-8218-1020-0 (1960). Chapter VII:'Interpolation by Polynomials'.
Text derived from https://en.wikipedia.org/w/index.php?oldid=1360994511 under CC BY-SA 4.0; modifications have been made.