Hyperbolic wavelet approximation (Q1381491)

Let \(D\) be a set of dyadic intervals of the from \(I_k=[j2^{-k},(j+1)2^{-k}]\), \(j,k\in{\mathbb{Z}}\), let \(\psi\) be a univariate wavelet, and denote \(\psi_I=2^{k/2}\psi(2^k\cdot-j)\), then \(\psi_I(z_1,\ldots,z_d)=\psi_{I_1}(z_1)\cdots\psi_{I_d}(z_d)\) for \(I=I_1\times\cdots\times I_d\in D^d\), is a multivariate tensor product wavelet basis in \({\mathbb{R}}^d\). The system \({\mathcal H}_n=\text{cl span}\{\psi_I: | I| >2^{-n}\}\) where cl is the closure in \(L_p=L_p({\mathbb{R}}^d)\) is called a hyperbolic wavelet system. This paper characterizes those functions in \(L_p\) (\(0<p\leq\infty\)) which can be approximated by functions from \({\mathcal H}_n\) with a given order of approximation. To this end, generalizations of the Jackson and Bernstein inequalities are proved under several conditions for \(\psi\). The proof relies on Littlewood-Paley theory. For similar results for hyperbolic trigonometric systems see [\textit{R. DeVore, B. Jawerth}, and \textit{V. Popov}, Am. J. Math. 114, No. 4, 737-785 (1992; Zbl 0764.41024)].