Tractability of infinite-dimensional integration in the worst case and randomized settings (Q647921): Difference between revisions

The authors consider the approximation of integrals of functions \(f\) with infinitely many variables in worst case and randomized settings. Here the class of \(\infty\)-variate functions \(f\) is build as an infinite sum of weighted tensor products of a reproducing kernel Hilbert space of scalar functions. The authors extend results from \textit{F. Y. Kuo, I. H. Sloan, G. W. Wasilkowski} and \textit{H. Woźniakowski} [J. Complexity 26, No.~5, 422--454 (2010; Zbl 1203.65057)] to more general function spaces. Mainly, the complexity and polynomial tractability of approximating integrals of such \(\infty\)-variate functions is studied. Polynomial tractability means that the \(\varepsilon\)-complexity (that is the minimal cost of all algorithms whose worst case errors are \(\leq \varepsilon\)) is \({\mathcal O}(\varepsilon^{-p})\), where the smallest \(p\leq 0\) is the exponent of tractability. The authors determine upper bounds of the tractability exponent and show that some bounds are sharp.