Rate of convergence of the Pólya algorithm from polyhedral sets (Q2567267)

Let \(K\) be a closed convex subset of \({\mathbb R}^n\) and \({h\in {\mathbb R}^n\setminus K}\). Let \(h_p\), \({p\in (1,+\infty)}\), denote the best \({l_p}\)-approximation of \(h\) from \(K\). It is well-known that \(h_p\) is always unique. In general, the best uniform approximation is not unique, but there exists a unique so called ''strict uniform approximation'' \(h^*_{\infty}\) [see \textit{M. Marano}, Approximation Theory Appl. 6, No. 1, 99--109 (1990; Zbl 0729.41030)]. Under some restrictions on \(K\) we have \({\lim_{p\to\infty}h_p=h^*_{\infty}}\). This convergence is called the Pólya algorithm. In the paper under review some geometric property (called property \(A_\infty\)) of sets in \({\mathbb R}^n\) is defined. Any convex polyhedral set satisfies this property. It is proved that if \(K\) satisfies the property \(A_\infty\), then \[ \limsup_{p\to\infty} p\| h_p-h^*_{\infty}\| <\infty. \]