Tractability of multivariate approximation over a weighted unanchored Sobolev space (Q843721)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Tractability of multivariate approximation over a weighted unanchored Sobolev space
scientific article

    Statements

    Tractability of multivariate approximation over a weighted unanchored Sobolev space (English)
    0 references
    0 references
    0 references
    15 January 2010
    0 references
    Let \(I=[0,1],\) \(m\geq1\) and \(H^{m}(I)\) be the standard Sobolev space, endoved with the unusual inner product denoted by \((f,g)_{H_{1,m}}=\int_{0} ^{1}f(x)g(x)dx+\int_{0}^{1}f^{(m)}(x)g^{(m)}(x)dx,\) and the corresponding norm \(\|\;\|_{1,m}\). In this case, the space is denoted by \(H_{1,m}.\) For \(d\in\mathbb{N}\) one defines the reproducing kernel Hilbert space \(H_{d,m}=H_{1,m}^{\otimes d}\) as \(d\)-fold tensor product of \(H_{1,m}\) endowed with the induced inner product \[ (f,g)_{H_{d,m}}=\int_{I^{d} }f(x)g(x)dx+\sum_{_{\substack{{u\subseteq[ d]}\\ {u\neq\phi}}}} \int_{I^{d}}\frac{\partial^{m|u|}}{\partial^{m}x_{u} } f(x)\frac{\partial^{m|u|}}{\partial^{m}x_{u} }g(x)dx,\;f,g\in H_{d,m}, \] where \(|u|\) denotes the size of \(u\subseteq[d]:=\{1,2,\dots,d\}\) and \(x_{u}\) denotes the vector whose components are those component \(x_{j}\) of \(x\) for which \(j\in u\). Also, let \(\Gamma=\{\gamma_{d,u}\geq0:\) nonenpty \(u\subseteq[ d]\), \(d\in \mathbb{N}\}\) be a given set of weights. The authors consider functions \(f\) belonging to the unit ball of the the space \(H_{d,m,\Gamma}\) so that \[ \int_{I^{d}}f^{2}(x)dx+\sum_{_{\substack{{ u\subseteq[ d]}\\{u\neq\phi}}}} \frac{1}{\gamma_{d,u}}\int_{I^{d}}\left( \frac{\partial ^{m|u|}}{\partial^{m}x_{u}}f(x)\right) ^{2}dx\leq1 \] and study the \(L_{2}(I^{d})\) approximation of such a function, when the approximation is given by an algorithm \(A_{d,n}\) using at most \(n\) information operations from a class \(\wedge\) of linear functionals on \(H_{d,m,\Gamma}.\) The worst case error of \(A_{d,n}\) is given by \(e(A_{d,n};\wedge)=\sup \{\|f-A_{d,n}f\|_{L^{2}(I^{d})}:f\in\mathcal{B} H_{d,m,\Gamma}\}\) and \(n\)th minimal error is defined as \(e(n,\wedge )=\inf\{e(A_{d,n};\wedge):A_{d,n}\}.\) The \(n\)th minimal error is of order \(n^{-(m-\delta)}\) for any \(\delta>0.\) Some types of tractability are considered and discussed using this order of minimal error.
    0 references
    0 references
    multivariate approximation
    0 references
    unachored Sobolev space
    0 references
    tractability
    0 references
    0 references