On the best \(L^ p\) multipoint local approximation (Q1089551)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On the best \(L^ p\) multipoint local approximation |
scientific article |
Statements
On the best \(L^ p\) multipoint local approximation (English)
0 references
1985
0 references
Let \(x_ 1<x_ 2<...<x_ k\) be k points on the real line. For \(\epsilon >0\), let \(I_{\epsilon}\) be the disjoint union of the intervals centered at \(x_ i\) and of length \(2\epsilon\). For a function \(f\in L^ p(I_{\epsilon})\) use the following notation \(\| f\|_{\epsilon}=(\int_{I_{\epsilon}}| f(t)|^ p(1/| I_{\epsilon}|)dt)^{1/p},\) where \(| I_{\epsilon}| =2\epsilon k\). For \(f\in L^ p\), if there exist numbers \(A=A(x_ 0)\), \(B=B(x_ 0)\) such that \(\int_{| t| \leq \epsilon}| f(x_ 0+t)-A-Bt|^ pdt=o(\epsilon^{p+1})\) as \(\epsilon\to 0\), then say that f has a derivative at \(x_ 0\) in \(L^ p\) sense, and write \(f\in t_ 1^ p(x_ 0)\). The authors prove the following theorem: Let \(1<p<\infty\) and \(f\in t^ p_ 1(x_ i)\), \(i=1,2,...,k\). Assume \(k\leq n\leq 2k-2\), and for each \(\epsilon >0\) let \(P_{\epsilon}(f)\) be the best \(L^ p\) approximation of f on \(I_{\epsilon}\) from \(\pi^ n\), which denotes the linear space of all algebraic polynomials with degrees not exceeding n (i.e. \(P_{\epsilon}(f)\) satisfies \(\| f- P_{\epsilon}(f)\|_{\epsilon}=\inf_{P\in \pi^ n}\| f- P\|_{\epsilon}).\) Then \(P_{\epsilon}(f)\) converges coefficientwise to a polynomial \(P_ 0(f)\in \pi^ n\) as \(\epsilon\to 0\). Furthermore, \(P_ 0(f)\) is the unique polynomial in \(\tilde H_ 1\), which minimizes \(\sum^{k}_{i=1}| B(x_ i)-P'(x_ i)|^ p\) over all \(P\in \tilde H_ 1\), where \(\tilde H_ 1=\{P\in \pi^ n:P(x_ i)=A(x_ i)\), \(i=1,...,k\}\).
0 references
local best uniform approximation
0 references
Taylor polynomial
0 references