Adaption helps for some nonconvex classes (Q579828)

The problem of numerical integration of functions from a given class F is considered. It is shown that if F is convex and balanced then adaptive quadrature algorithms are not better than nonadaptive ones. Examples of nonconvex classes of functions F for which adaption is substantially better than nonadaption are presented. From a practical point of view the adaptive algorithms constructed may not be suitable because they require knowledge of parameters, difficult to specify a priori.