The best \(m\)-term approximations on generalized Besov classes \(M\, B_{q, \theta}^{\Omega}\) with regard to orthogonal dictionaries (Q619046)

The setting of this paper is as follows: Let \({\mathcal D}\) be a dictionary in a Banach space \(X\), (i.e., a system of functions with domain \(X\)), and let the best \(m\)-term approximation with regard to \({\mathcal D}\) be defined by \[ \sigma_n(f, {\mathcal D})_X:=\inf \left\{ \left\| f - \sum_{i=1}^m c_i g_i \right\|\right\}, \] where the infimum is taken over the set of all possible linear combinations of \(m\) functions \(g_i\) in \({\mathcal D}\). For a function class \(F \subset X\) and a collection \(\mathbf D\) of dictionaries, the quantity \(\sigma_n(F, {\mathbf D})_X:=\inf_{{\mathcal D} \in {\mathbf D}} \sup_{f \in F} \sigma_n(f, {\mathcal D})_X\) is the lowest bound of the best \(m\)-term approximation of a given function class \(F\) with regard to any dictionary \({\mathcal D} \in {\mathbf D}\). In this paper, the author investigates \(m\)-term approximation with regard to orthogonal dictionaries. He considers the problem in the periodic multivariate case for generalized Besov classes \(M B^{\Omega}_{q, \theta}\) under the condition \(\Omega(t)=\omega(t_1 \cdot \cdots \cdot t_d)\), where \(\omega(t) \in {\mathbf \Psi}^{\ast}_{\ell}\) is a univariate function, and proves that the well-known dictionary \(U^d\) which consists of trigonometric polynomials (shifts of the Dirichlet kernels) is nearly optimal among orthogonal dictionaries. Moreover, he establishes that for these classes near-best \(m\)-term approximation, with regard to \(U^d\), can be achieved by simple greedy-type algorithms. For related work see e.g. \textit{P. Oswald} [J. Fourier Anal. Appl. 7, No. 4, 325--341 (2001; Zbl 1003.41020)]