Sampling recovery on function classes with a structural condition

Let \(X\) be a Banach space with a norm \(\|\cdot\|_X\) and \(\mathcal D=\{g_j\}^\infty_{j=1}\subset X\). For \(J\subset \mathbb N=\{1,2,\dots\}\) let \(V_j(\mathcal D)=span\{g_j: j\in J\}\) and for \(v\in\mathbb N\) let \(\chi_V(\mathcal D)\) be the set of all subspaces \(V_j(\mathcal D)\) with \(|J|=v\), while \(\Sigma_v(\mathcal D)=\cup_{V\in \chi_v(\mathcal D)}V\). Then for \(f\in X\) \N\[\N\sigma_v(f,\mathcal D)_X=\inf\limits_{g\in\Sigma_v(\mathcal D)}\|f-g\|_X, \quad v\in\mathbb N. \N\]\NLet \N\[\Nf=\sum_{\mathbf{k}\in\mathbb Z^d}a_{\mathbf{k}}(f)\psi_{\mathbf{k}}, \quad \sum_{\mathbf{k}\in\mathbb Z^d}|a_{\mathbf{k}}(f)|<\infty, \tag{1}\N\]\Nand for \(\mathbf{s}=(s_1,s_2,\dots,s_d)\) with nonnegative integer components \(\rho(\mathbf{s})= \{\mathbf{k}\in\mathbb Z^d:[2^{s_j-1}]\leq |k_j|<2^{s_j}\}\), \(\delta_{\mathbf{s}}(f,\Psi)=\sum_{\mathbf{k}\in \rho(\mathbf{s})}a_{\mathbf{k}}(f)\psi_{\mathbf{k}}\), \(f_j=\sum_{\|\mathbf{s}\|_{l^1}=j}\delta_{\mathbf{s}}(f,\Psi)\) and \(\|f\|_{A_\beta(\Psi)}=\left(\sum_{\mathbf{k}\in\mathbb Z^d}|a_{\mathbf{k}}(f)|^{\beta}\right)^{1/\beta}\), \(\beta\in (0,1]\). For parameters \(a\in [0,+\infty)\), \(b\in\mathbb R\) we consider the set \(W^{a,b}_{A_\beta}(\Psi)\) of functions with representations (1) and satisfying \(\|f_j\|_{A_\beta(\Psi)}\leq 2^{-aj}(\max(j,1))^{(d-1)b}\). Finally, by \(\Psi(Q)\), \(Q\subset \mathbb Z^d\), we denote the space of \(f\) of type (1) with spectrum in \(Q\).\N\NThe main result of the paper is\N\N\textbf{Theorem}. Let \(1<p<\infty\), \(a>0\), \(b\in\mathbb R\), \(\Omega\) be a compact set in \(\mathbb R^d\), \(\mu\) be a probability measure. Suppose that \(\|\psi_{\mathbf{k}}\|_{L_p(\Omega,,\mu)}\leq 1\). Then there exist \(c\) and \(C\) such that, for given \(v\in\mathbb N\), there exists a set \(Q\subset\mathbb Z^d\), \(|Q|\leq v^{c}\), independent of \(\mu\) with the following property: for every \(f\in W^{a,b}_{A_\beta}(\Psi)\) \N\[\N\sigma_v(f, \Psi(Q))_{_{L_p(\Omega,,\mu)}}\leq Cv^{-a+1/p^*-1/\beta}(\log(2v))^{(d-1)(a+b)}, \quad p^*=\max(p,2).\N\]

Cites work

Cited in

(3)

This page was built for publication: Sampling recovery on function classes with a structural condition

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6970161)