Fast Evaluation and Interpolation at the Chebyshev Sets of Points
نویسنده
چکیده
Stable polynomial evaluation and interpolation at n Chebyshev or adjusted (expanded) Chebyshev points is performed using O(nlog’ n) arithmetic operations, to be compared with customary algorithms either using on the order of n* operations or being unstable. We also evaluate a polynomial of degree d at the sets of n Chebyshev or adjusted (expanded) Chebyshev points using O(dlog d log n) if n 5 d or O((d log d + n) log d) arithmetic operations ifn>d. 1. INTR~OUCTI~N Consider the set of Fourier points on the unit circle in the complex plane, {W 2a+1 , k=O,l,..., n-l}, (I) w = exp ( > e being a primitive (4n)-th root of 1, and also the Chebyshev set of the real coordinates of these points, {Xk =,,,(yT), k=O,l,... ,n--1) ([41, I519 [71). Fast F ourier transform (FFT) is a very effective means of interpolation to a function at the Fourier set (1) by a polynomial and of the evaluation of a polynomial at such a set of points. We will present fast stable algorithms for the evaluation and interpolation at the Chebyshev set (2), which are almost as efficient as fast Fourier transform (FFT) at the set (1), provided that n + 1 = 2h is an integer power of 2; those computations at the Chebyshev set (2) are highly important for the approximation to functions by polynomials and for stable polynomial evaluation (see Section 4 for our comments on the computational cost and for a discussion). The results apply also to the sets of adjusted (expanded) Chebyshev points {Yk = (‘xk + b, k=O,l,... ,n1) for two real constants a and b. Hereafter all logarithms are to the base 2. 2. EVALUATION Next we will extend the known scheme for FFT to polynomial evaluation at the set (2). Given a polynomial P(x) = Cid,c Pixi, we write P(x) = Pc(x2) + xPr(x2), Pc(x*) = Cj P2j X2j , pI(X2>= Cj pZj+lX 2j, Qh(y) = Ph (9) for j ranging from 0 to [$I, h=O,l, y= 1 2x2, so that P(x) = Qdy) + TQI(Y), Qh(y) = h(z2), Y = 1 2x2, h = O,l. (3) Those equations reduce the evaluation of P(x) at the n point x set (2) to the evaluation of two smaller degree (at most half-degree) polynomials Qc(y), Qi(y) at the Chebyshev 4 points y-set This research has been supported by NSF Grant CCR-8805782
منابع مشابه
Multivariate polynomial interpolation on Lissajous-Chebyshev nodes
In this contribution, we study multivariate polynomial interpolation and quadrature rules on non-tensor product node sets linked to Lissajous curves and Chebyshev varieties. After classifying multivariate Lissajous curves and the interpolation nodes related to these curves, we derive a discrete orthogonality structure on these node sets. Using this discrete orthogonality structure, we can deriv...
متن کاملRational Interpolation at Chebyshev points
The Lanczos method and its variants can be used to solve eeciently the rational interpolation problem. In this paper we present a suitable fast modiication of a general look-ahed version of the Lanczos process in order to deal with polynomials expressed in the Chebyshev orthogonal basis. The proposed approach is particularly suited for rational interpolation at Chebyshev points, that is, at the...
متن کاملNewton Interpolation in Fejér and Chebyshev Points
Let T be a Jordan curve in the complex plane, and let Í) be the compact set bounded by T. Let / denote a function analytic on O. We consider the approximation of / on fî by a polynomial p of degree less than n that interpolates / in n points on T. A convenient way to compute such a polynomial is provided by the Newton interpolation formula. This formula allows the addition of one interpolation ...
متن کاملAn Algorithm Based on the Fft for a Generalized Chebyshev Interpolation
An algorithm for a generalized Chebyshev interpolation procedure, increasing the number of sample points more moderately than doubling, is presented. The FFT for a real sequence is incorporated into the algorithm to enhance its efficiency. Numerical comparison with other existing algorithms is given.
متن کاملError bounds for approximation in Chebyshev points
This paper improves error bounds for Gauss, Clenshaw-Curtis and Fejér’s first quadrature by using new error estimates for polynomial interpolation in Chebyshev points. We also derive convergence rates of Chebyshev interpolation polynomials of the first and second kind for numerical evaluation of highly oscillatory integrals. Preliminary numerical results show that the improved error bounds are ...
متن کامل