The best \(m\)-term approximation with respect to polynomials with constant coefficients (Q682149): Difference between revisions

Let \(X\) be an infinite-dimensional Banach space and \({\mathcal B}=(e_n)_{n=1}^\infty\) a normalized Schauder basis of \(X\) with biorthogonal functionals \((e^*_n)_{n=1}^\infty\). For each set \(A\subset \mathbb{N}\) with card\((A) =| A| <\infty\), put \(1_A=\sum_{j\in A}e_j\) and \(P_A(x)= \sum_{n\in A}e^*_n(x)e_n\). Also, for a given collection \(\;(\eta_j)_{j\in A}\in\{\pm 1\}\) of signs, put \(1_{\eta A}=\sum_{j\in A}\eta_je_j\) and let \([1_{\eta A}] \) and \([e_n,n\in A]\) be the subspaces generated by \(1_{\eta A}\) and \(\{e_n,n\in A\}\), respectively. Then the expression \(\mathcal G_m= \sum_{j=1}^me^*_{\rho(j)}(x)e_{\rho(j)}\), where \(\rho\) is a greedy ordering, is called the \(m\)-th greedy approximant of \(x\), see [\textit{S. V. Konyagin} and \textit{V. N. Temlyakov}, East J. Approx. 5, No. 3, 365--379 (1999; Zbl 1084.46509)], and \(\sigma_m:=\inf\{d(x,[e_n,n\in A]) : A\subset \mathbb{N},\, | A| =m\}\) is called the \(m\)-term approximation with respect to \(\mathcal B\). A basis \(\mathcal B\) satisfying the condition \[ \| x-\mathcal G_m(x)\| \leq C \sigma_m(x)\quad\text{ for all } x\in X \text{ and }m\in\mathbb{N},\tag{\(*\)} \] where \(C\) is an absolute constant, is called a greedy basis. The authors introduced in [J. Approx. Theory 215, 28--39 (2017; Zbl 1369.46014)] the best \(m\)-term approximation with respect to polynomials with constant coefficients by the formula \[ \mathcal D^*_m:=\inf\{d(x,[1_{\eta A}]) : A\subset \mathbb{N},\, | A| =m,\, (\eta_n)\in\{\pm 1\}\} \] and obtained a new description of greedy bases in these terms. In this paper, they give a new direct proof of the equivalence between the conditions \((*)\) and \[ \| x-\mathcal G_m(x)\| \leq C \mathcal D^*_m(x)\quad \text{ for all } x\in X \text{ and }m\in\mathbb{N}, \] even in the more general case of \((t,w)\)-greedy bases.