Optimal recovery of operators in sequence spaces (Q6970949)

This paper is on the optimal recovery of operators in sequence spaces. The results of this paper are closely related to the paper in [\textit{V. F. Babenko} et al., Mat. Stud. 56, No. 2, 193--207 (2021; Zbl 1502.41008)].\N\NSpecifically, for a given non-increasing sequence \(t=\{t_k\}_{k=1}^\infty\) of non-negative numbers vanishing at infinity, consider \(T:\ell_q\rightarrow \ell_q\) with \(Th:=\{t_kh_k\}_{k=1}^\infty\) for \(h\in\ell_q\), the paper considers the class\N\[\NW_q^T:=\{x=Th: h\in\ell_q, \|h\|_{\ell_q}\le 1\}\N\]\Nand the operator\N\[\NA_\alpha:\ell_q \rightarrow \ell_q\mbox{ with } A_\alpha x =(\alpha_1 x_1,\alpha_2 x_2,\ldots),\N\]\Nwhere \(\alpha=\{\alpha_k\}_{k=1}^\infty\) being a sequence of non-negative numbers such that \(\tau=\{\tau_k=\alpha_k t_k\}_{k=1}^\infty\) is non-increasing and vanishing at infinity. Let \(I: W_q^T\rightarrow \mathfrak{B}(W_q^T)\) be a mapping called information that maps \(x\in W_q^T\) to a certain non-empty subset \(I(x)\). An arbitrary mapping \(\Phi: I(x)\rightarrow \ell_q\) is called a method of recovery of the operator \(A_\alpha\). The error of the method of the recovery \(\Phi\) of the operator \(A_\alpha\) on the class \(W_q^T\) given information \(I\) is defined by\N\[\N\mathcal{E}(A_\alpha,W_q^T, I,\Phi):=\sup_{x\in W_q^T}\sup_{z\in I(x)}\|A_\alpha x - \Phi(z)\|_{\ell_q}.\N\]\NThe quantity\N\[\N\mathcal{E}(A_\alpha,W_q^T, I):=\inf_{\Phi}\mathcal{E}(A_\alpha,W_q^T, I,\Phi)\N\]\Nis called the error of optimal recovery of operator \(A_\alpha\) on the class \(W_q^T\) given information \(I\) and the method \(\Phi^*=\arg\inf_{\Phi}\mathcal{E}(A_\alpha,W_q^T, I,\Phi)\) is called optimal.\N\NThe paper considers three cases of different \(I\):\N\begin{itemize}\N\item[(1)] \(Ix = I_{\bar\varepsilon}^nx= (x_1,\ldots,x_n)+[-\varepsilon_1,\varepsilon_1]\times\cdots\times[-\varepsilon_n,\varepsilon_n]\);\N\item[(2)] \(Ix = I_{\varepsilon,p}^nx= (x_1,\ldots,x_n)+B[\varepsilon,\ell_p^n]\), where \(B[\varepsilon,\ell_p^n]\) is the ball of radius \(\varepsilon\) in the space \(\ell_p^n\) of finite sequences.\N\item[(3)] \(Ix = I_{\varepsilon,p}x= x+B[\varepsilon,\ell_p]\), where \(B[\varepsilon,\ell_p]\) is the ball of radius \(\varepsilon\) in the space \(\ell_p\).\N\end{itemize}\NThe method of recovery \(\Phi_m^*: \ell_p\rightarrow \ell_q\) is defined by\N\[\N\Phi_m^*(a) = \{a_1\alpha_1(1-\frac{\tau_{m+1}^q}{\tau_1^q}),\ldots,a_m\alpha_m(1-\frac{\tau_{m+1}^q}{\tau_m^q}) ,0,\ldots\}, a\in\ell_p.\N\]\NFor each of the above three cases, the paper shows that under certain conditions, \(\Phi_m^*\) is optimal.\N\NFinally, the paper consider the case for \(H\) a Hilbert space with orthonomral basis \(\{\varphi_n\}_{n=1}^\infty\), the operator \(\mathcal{A}_\alpha:H\rightarrow H\) on the class \(\mathcal{W}^T:=\{x=\sum_nt_n c_n\varphi_n: \sum_n|c_n|^2\le1\}\) given by\N\[\N\mathcal{A}_\alpha x = \sum_n \alpha_n x_n \varphi_n\mbox{ with } x = \sum_n x_n\varphi_n\in H,\N\]\Nand the information operator \(\mathcal{I}_{p,\varepsilon}: H\rightarrow \ell_p\) with \(\mathcal{I}_{p,\varepsilon} x = (x_1,x_2\ldots)+ B[\varepsilon,\ell_p]\) for \(x=\sum_n x_n\varphi_n\). Due to the isomorphism between \(\ell_2\) and \(H\), one has\N\[\N\mathcal{E}(\mathcal{A}_\alpha,\mathcal{W}^T, \mathcal{I}_{\varepsilon,p})=\mathcal{E}(A_\alpha,W_2^T, I_{\varepsilon,p}).\N\]\NMoreover, the methods of recovery \(F_{m,\lambda}^*:=\mathfrak{U}\circ \Phi_{m,\lambda}^*\) are optimal with \(\mathfrak{U}: \ell_2\rightarrow H\) given by \(\mathfrak{U}(x_1,x_2,\ldots) = \sum_n x_n\varphi_n\). The special case \(W_2^\mu(\mathbb{T})\) of functions having \(L_2\)-bounded Weyl derivatives of order \(\mu\) and the Weyl fractional differentiation operator \(\mathcal{A}_\alpha = \frac{d^\gamma}{dx^\gamma}\) is studied in the last part of the paper.