Representation and approximation of multivariate functions with mixed smoothness by hyperbolic wavelets (Q1433372)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Representation and approximation of multivariate functions with mixed smoothness by hyperbolic wavelets |
scientific article |
Statements
Representation and approximation of multivariate functions with mixed smoothness by hyperbolic wavelets (English)
0 references
15 June 2004
0 references
Let \(\psi \) be a univariate wavelet and let \(\psi _{j,k}( u) =2^{k/2}\psi ( 2^{k}u-j) \), \(k,j\in {\mathbb Z}\), the induced orthonormal basis. One can lift this system to a basis to \(d\) dimensions by forming tensor products \(\psi _{j,k}=\psi _{j_{1},k_{1}}\otimes \dots \otimes \psi _{j_{d},k_{d}}\) where \(k=( k_{1},\dots ,k_{d}) \) and \(j=( j_{1},\dots ,j_{d}) \in {\mathbb Z}^{d}\). The \(L^{p}\)-closure of the linear span generated by the functions \(\psi _{j,k},j,k\in {\mathbb Z}^{d}\), such that \(k_{1}\gamma _{1}+\dots +k_{d}\gamma _{d}\leq n\) is denoted by \( H_{n}^{\gamma }\) where \(\gamma =( \gamma _{1},\dots ,\gamma _{d}) \in {\mathbb R}^{d}\) and \(n\) is a natural number. Such systems are called hyperbolic systems and have been investigated by \textit{R. A. DeVore}, \textit{S. V. Konyagin} and \textit{V. N. Temlyakov} in [Constructive Approximation 14, No. 1, 1--26 (1998; Zbl 0895.41016)]. If \(\psi \) has compact support and \( \gamma _{1}=\dots =\gamma _{d}=1\) this condition means that each \(\psi _{j,k}\in H_{n}^{\gamma }\) has a volume of support \(\geq \) constant \(\cdot 2^{-n}.\) Similarly one defines the space \(H_{n}^{\gamma ,\ast }\) with respect to the orthonormal basis \(\psi _{j,k}^{\ast }( x) =\psi _{j_{1},k_{1}}^{\ast }\otimes \dots \otimes \psi _{j_{d},k_{d}}^{\ast }\), \( j\in {\mathbb Z}^{d}\), \(k\in {\mathbb N}_{0}^{d},\) where \(\psi _{j,k}^{\ast }:=\psi _{j,k}\) for \(k>0\) and \(\psi _{j,0}^{\ast }( u) :=2^{1/2}\varphi ( 2u-j) \) for all \(j\in {\mathbb Z}\); here \(\varphi \) is the scaling function of a multiresolution analysis. The author considers Sobolev-type, Hölder-type and Besov-type spaces with mixed smoothness conditions up to the order \(r=( r_{1},\dots ,r_{d}) \) for multivariate functions \(f:{\mathbb R}^{d}\to {\mathbb C}.\) The main result of the paper is the equivalence of the norms in these function classes to norms defined in terms of the wavelet coefficients of \(\psi _{j,k}^{\ast },\) \(j\in {\mathbb Z}^{d}\) and \(k\in {\mathbb N}_{0}^{d};\) as usual it is assumed that \(r_{i}\) for \( i=1,\dots ,d\) is smaller than the order of regularity of \(\psi \) and \(\varphi .\) The second main result gives asymptotic estimates for the best approximation in the above function spaces with respect to the subspaces \( H_{n}^{\gamma ,\ast }\) and \(H_{n}^{\gamma }.\)
0 references
hyperbolic wavelets
0 references
functions with mixed smoothness
0 references
decomposition of functions
0 references
best approximation
0 references
0 references