Bounding the Lebesgue constant for Berrut's rational interpolant at general nodes (Q387252)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Bounding the Lebesgue constant for Berrut's rational interpolant at general nodes
scientific article

    Statements

    Bounding the Lebesgue constant for Berrut's rational interpolant at general nodes (English)
    0 references
    0 references
    0 references
    0 references
    0 references
    20 December 2013
    0 references
    The authors study interpolations with \(n+1\) distinct real nodes in the interval \([a,b]\) given by \(X_n=\{x_0,x_1,\dots,x_n\}\) and the corresponding set of cardinal basis functions \(B_n=\{b_0,b_1,\dots,b_n\}\) (i.e. \(b_j(x_k)=\delta_{jk}\)). The Lebesgue constant is the norm of the operator mapping \(f\in{\mathcal C}^0[a,b]\) to \(g=\sum_{j=0}^n\,b_jf(x_j)\) and given by \[ \Lambda_n(X_n,B_n)=\max_{a\leq x\leq b}\,\sum_{j=0}^n\,|b_j(x)|. \] The specific type of interpolant used in this paper is Berrut's first rational interpolant with basis functions \[ b_i(x)={(-1)^i\over x-x_i}/ \sum_{j=0}^n\,{(-1)^j\over x-x_j},\quad i=0,\dots,n, \] and the interest of the authors focuses on the growth of the corresponding Lebesgue constant \[ \Lambda(X_n)=\max_{a\leq x\leq b}\,{ \sum_{j=0}^n\,{1\over |x-x_j|}\over\left| \sum_{j=0}^n\,{(-1)^j\over x-x_j}\right|}. \] The main result is { Theorem 2.2.} If \((X_n)_{n\in\mathbf{N}}\) is a family of well-spaced nodes, then there exists a constant \(c>0\) such that \[ \Lambda(X_n)\leq c\ln{n}\text{ for any }n\geq 2. \] The concept of a well-spaced family is introduced by the requirement of the existence of constants \(C,R\geq 1\), independent of \(n\), such that each of the sets of nodes \(X_n\) satisfies \[ \begin{aligned} &{x_{k+1}-x_k\over x_{k+1}-x_j}\leq {C\over k+1-j}\;(0\leq j\leq k,\,0\leq k\leq n-1),\\& {x_{k+1}-x_k\over x_j-x_k}\leq {C\over j-k}\;(k+1\leq j\leq n,\,0\leq k\leq n-1),\\& {1\over R}\leq {x_{k+1}-x_k\over x_k-x_{k-1}}\;(1\leq k\leq n-1).\end{aligned} \] Furthermore, the authors give a method to construct well-spaced families using regular distribution functions: - distribution function: strictly increasing bijection on the interval \([0,1]\), - regular: \(F\in{\mathcal C}^1[0,1],\;F'\) has a finite number of zeros \(T=[t_1,t_2,\dots,t_{\ell}]\subset [0,1]\) with finite multiplicities and there exist positive real numbers \(r_1,r_2,\dots,r_{\ell}\) with \[ G_j(t)=F'(t)/|t-t_j|^{r_j}\text{ is continuous with }\lim_{t\rightarrow t_j}\;G_j(t)>0\text{ for }j=1,\dots,\ell. \] \(X_n=X_n(F)\) is then defined by \(x_k=F(k/n)\), \(0\leq k\leq n\).
    0 references
    barycentric rational approximation
    0 references
    Lebesgue constant
    0 references

    Identifiers