On the Lebesgue constant of barycentric rational interpolation at equidistant nodes (Q443847): Difference between revisions

The setting of the paper is approximation of a function \(f:[a,b]\rightarrow\mathbb R\) by some function taken from a finite-dimensional linear subspace of the Banach space \(C^0[a,b]\) of continuous functions on \([a,b]\), equipped with the maximum norm, such that \(g\) interpolates \(f\) at \(n+1\) distinct points \(a=x_0<x_1<\cdots<x_n=b\). \vskip0.2cm The interpolants used are the so-called \textit{barycentric rational interpolants} due to \textit{M. S. Floater} and \textit{K. Hormann} [Numer. Math. 107, No. 2, 315--331 (2007; Zbl 1221.41002)], given by \[ b_j(x)=\left.{(-1)^j \beta_j\over x-x_j}\right/\sum_{i=0}^n\,{(-1)^i\beta_i\over x-x_i},\;j=0,\dots,n \] where for \(n\geq 2d\) equidistant nodes the weights are explicitly given by \[ \beta_j=\sum_{k=d}^n\,{d\choose k-j}. \] Important quantities are now {\parindent=6mm \begin{itemize}\item[-] the \textit{Lebesgue function} \[ \Lambda_n(x)=\sum_{j=0}^n\,|b_j(x)|, \] \item[-] the \textit{Lebesgue constant} \[ \Lambda_n=\max_{a\leq x\leq b}\,\Lambda_n(x). \] \end{itemize}} In case of equidistant nodes, the properties of barycentric rational interpolation depend on the constant distance \(h\) between the nodes only, therefore (without loss of generality) it can be assumed that the interval is \([0,1]\) and that the nodes are given by \(x_j=j/n,\;(0\leq j\leq n)\). The main results of the paper are then Theorem 1: \[ \Lambda_n\leq 2^{d-1}(2+\ln{n})\;(d\geq 1). \] Theorem 2: \[ \Lambda_n\geq {1\over 2^{d+2}}\,{2d+1\choose d}\ln{({n\over d}-1)}\;(d\geq 1), \] Proposition 1: If \(d=1\), then \[ \Lambda_n\geq a_n\ln{(n)}+b_n,\text{ where} \lim_{n\rightarrow\infty}\,a_n={2\over \pi},\quad \lim_{n\rightarrow\infty}\,b_n=0. \] Finally, the paper shows the graphs of several numerical experiments.