On optimal recovery in \(L_2\) (Q2041064)

Let \(\Omega\) be a compact subset of \(\mathbb{R}^d\) with the probability measure \(\mu\). For a fixed \(m\) and a set of points \(\xi := \{\xi^j\}\subset\Omega\), let \(\Phi_{\xi}\) be a linear operator from \(\mathbb{C}^m\) into \(L_2(\Omega,\mu)\). For a subset \(F\) of \(L_2(\Omega,\mu)\), let \( \rho_m(F;L_2):=\inf\limits_{\Phi_{\xi},\xi}d(F,\Phi_{\xi},\xi;L_2), \) where \[ d(F,\Phi_{\xi},\xi;L_2):=\sup_{f\in F}\|f - \Phi_{\xi}(f(\xi^1),\dots,f(\xi^m)) \|_{L_2(\Omega,\mu)}. \] The main result is the following Theorem. Let \(F\) be a centrally symmetric compact of \(C(\Omega)\). There exist two positive absolute constants \(b\) and \(B\) such that \[ \rho_{bn}(F;L_2) \leq B d_n(F;L_{\infty}), \] where \(d_n\) is the Kolmogorov \(n\)-width. The proof uses discretization theorems based on some recent deep results related to the famous Kadison-Singer problem. The author demonstrates on a number of examples of \(F\) from classes of functions with mixed smoothness that the obtained inequality provides a powerful tool for estimating errors of optimal recovery.