Uniqueness of best \(\varphi\)-approximation from the set of splines with infinitely many simple knots (Q1433347)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Uniqueness of best \(\varphi\)-approximation from the set of splines with infinitely many simple knots
scientific article

    Statements

    Uniqueness of best \(\varphi\)-approximation from the set of splines with infinitely many simple knots (English)
    0 references
    0 references
    0 references
    15 June 2004
    0 references
    Let \(\phi\) be a convex function on \([0,\infty)\) such that \(\phi(0)=0\) , \(\phi(y)>0\) for \(y >0\) and for some \(y_0,c >0\), \(\phi(2y) \leq c \phi(y)\) for \(y \geq y_0\). For \(-\infty \leq a \leq b \leq \infty\), let \(J=(a,b)\) and \(\pi=\{x_i\}_{i\in \Gamma}\) with \(a < x_i < x_j <b\) whenever \(i <j\), where \(\Gamma =\mathbb Z\) or \(\mathbb N\) or \(-\mathbb N\). Let \(\varphi_\pi=\{\)piecewise polynomial functions \(g\) of degree at most \(n-1\) with simple knots in \(\pi\}\). Let \({\mathcal L}_\phi=\{h : h\) Lebesgue measurable on \(J\) and \(\int \phi(| h| ) < \infty\}\). In Section 2 of this very interesting paper, the authors prove the existence of best \(\phi\)-approximation from \(\varphi_\pi\) for any \(f \in {\mathcal L}_\phi\). A linear subspace \(\varphi\) of \(C(J)\) is said to have property \(A\), if for any \(U^\ast \in C(J)\) with \(| U^\ast| =| U| \) of some \(0 \neq U \in \varphi\) there exists a \(0 \neq h_0 \in \varphi\) such that \(h_0=0\) a.e. on Ker\,\(U^\ast\) and \(h_0U^\ast \geq 0\) on \(J\). Under the assumption that \(\varphi_\pi \cap {\mathcal L}_\varphi\) satisfies property \(A\), the authors show in Section 3 that for a continuous \(f\) in \({\mathcal L}_\phi\) the best \(\phi\)-approximation from \(\varphi_\pi\) is unique. In Section 4 the authors establish the main results of the paper, that both \(\varphi_\pi\) and \(\varphi_\pi \cap {\mathcal L}_\phi\) satisfy property \(A\).
    0 references
    0 references
    0 references
    0 references
    0 references
    unique best approximation
    0 references
    0 references