On the power of standard information for \(L_{\infty}\) approximation in the randomized setting (Q1035792)

Approximations from reproducing kernel Hilbert spaces are important in theory and practice, because these spaces provide vast and useful choices of approximating functions. In this paper, the approximation power from the spaces is studied in various cases. The cases include standard information consisting of function values of the approximand or alternatively of general linear information which means general linear functionals applied to the approximand. The different cases of approximation power are compared for the randomized setting and for the worst case setting. The approximations are always such that the error is measured in the Chebychev-norm.