English

Fast and stable approximation of analytic functions from equispaced samples via polynomial frames

Numerical Analysis 2022-03-08 v2 Numerical Analysis

Abstract

We consider approximating analytic functions on the interval [1,1][-1,1] from their values at a set of m+1m+1 equispaced nodes. A result of Platte, Trefethen \& Kuijlaars states that fast and stable approximation from equispaced samples is generally impossible. In particular, any method that converges exponentially fast must also be exponentially ill-conditioned. We prove a positive counterpart to this `impossibility' theorem. Our `possibility' theorem shows that there is a well-conditioned method that provides exponential decay of the error down to a finite, but user-controlled tolerance ϵ>0\epsilon > 0, which in practice can be chosen close to machine epsilon. The method is known as \textit{polynomial frame} approximation or \textit{polynomial extensions}. It uses algebraic polynomials of degree nn on an extended interval [γ,γ][-\gamma,\gamma], γ>1\gamma > 1, to construct an approximation on [1,1][-1,1] via a SVD-regularized least-squares fit. A key step in the proof of our main theorem is a new result on the maximal behaviour of a polynomial of degree nn on [1,1][-1,1] that is simultaneously bounded by one at a set of m+1m+1 equispaced nodes in [1,1][-1,1] and 1/ϵ1/\epsilon on the extended interval [γ,γ][-\gamma,\gamma]. We show that linear oversampling, i.e., m=cnlog(1/ϵ)/γ21m = c n \log(1/\epsilon) / \sqrt{\gamma^2-1}, is sufficient for uniform boundedness of any such polynomial on [1,1][-1,1]. This result aside, we also prove an extended impossibility theorem, which shows that such a possibility theorem (and consequently the method of polynomial frame approximation) is essentially optimal.

Keywords

Cite

@article{arxiv.2110.03755,
  title  = {Fast and stable approximation of analytic functions from equispaced samples via polynomial frames},
  author = {Ben Adcock and Alexei Shadrin},
  journal= {arXiv preprint arXiv:2110.03755},
  year   = {2022}
}
R2 v1 2026-06-24T06:43:14.339Z