Strong partially greedy bases and Lebesgue-type inequalities (Q2067508)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Strong partially greedy bases and Lebesgue-type inequalities |
scientific article |
Statements
Strong partially greedy bases and Lebesgue-type inequalities (English)
0 references
18 January 2022
0 references
The authors continue the study of Lebesgue-type inequalities for greedy algorithms. They consider a new concept denoted strong partially greedy Markushevich bases and analyze the Lebesgue-type inequalites associated to them. The setting is a semi-normalized Markushevich basis \(\mathcal B=(e_n)\) in a Banach space \(X\), that is \(X=\overline{[e_n:n\in \mathbb N]}\), with biorthogonal functionals \((e_n^*)\) such that \(e_n^*(x)=0\) for all \(n\) implies \(x=0\) and \(\sup\{\|e_n\|,\|e_n^*\|\}<\infty\) and they use the standard notation \(P_A(x)=\sum_{n\in A} e_n^*(x)e_n\) and \(P_m= P_{\{1,\cdots,m\}}\). As usual they denote by \(G_m\) the greedy operator \(G_m(x)=P_{A}(x)\) where \(A\) is a \(m\)-greedy set for \(x\), i.e., \(|A|=m\) and \(\min_{n\in A}|e_n^*(x)|\ge \max_{n\notin A}|e_n^*(x)|\). Comparing with the error terms \(\sigma_m(x)=\inf \{\|x-y\|: |\mathrm{supp} (y)|\le m\}\) or \(\tilde\sigma_m(x)= \inf \{\|x-P_B(x)\|: |B|\le m\}\) one can define the following Lebesgue-type constants \(L_m\) and \(\tilde L_m\) as the smallest values such that for all \(x\in X\) \[\|x-G_m(x)\|\le L_m \sigma_m(x), \quad \|x-G_m(x)\|\le \tilde L_m \tilde\sigma_m(x)\] is satisfied. Recall that the conditions \(L_m=O(1)\) and \(\tilde L_m=O(1)\) lead to the notion of greedy and almost-greedy basis, respectively. Similarly one can define \(L^{re}_m\) by the condition \( \|x-G_m(x)\|\le L^{re}_m \|x-P_m(x)\|\) for all \(x\in X\). Bases satisfying the condition \(L^{re}_m=O(1)\) are called partially greedy. The authors also consider the following variation denoted the \(m\)th-strong residual error \(\hat\sigma_m(x)=\inf_{0\le k\le m}\|x-P_k(x)\|\) and the new constant \(\hat L_m\) as the smallest value such that \(\|x-G_m(x)\|\le \hat L_m \hat\sigma_m(x)\) for all \(x\in X\). When \(\hat L_m=O(1)\) the \(M\)-basis is said to be \(C\)-strong partially greedy for \(C=\sup \hat L_m\). They present for semi-normalized Markushevich basis the extension of the result proved in the setting of \(w\)-partially greedy Schauder bases by \textit{S. J. Dilworth} et al. [Constr. Approx. 19, No. 4, 575--597 (2003; Zbl 1050.46011)] and also they characterize the \(1\)-strong partially greedy basis following the study by \textit{F. Albiac} and \textit{P. Wojtaszczyk} [J. Approx. Theory 138, No. 1, 65--86 (2006; Zbl 1087.65048)] and \textit{F. Albiac} and \textit{J. L. Ansorena} [Rev. Mat. Complut. 30, No. 1, 13--24 (2017; Zbl 1368.46015)] for \(1\)-greedy and \(1\)-almost-greedy bases.
0 references
nonlinear approximation
0 references
Lebesgue-type inequality
0 references
greedy algorithm
0 references
quasi-greedy bases
0 references
partially greedy bases
0 references