Automatic integration using asymptotically optimal adaptive simpson quadrature (Q495532)

The author derives a strategy for constructing an optimal (with respect to the computational cost) adaptive integration method for the integral \(\int_a^b f(x) dx\) based on Simpson's rule. The development is based on the assumption that \(f^{(4)}\) is bounded and strictly positive on \([a,b]\). In view of the well-known definiteness of Simpson's rule, this hypothesis allows to conclude that the sign of the quadrature error is the same over each subinterval. This property substantially simplifies the analysis and the optimality proof of the underlying subdivision strategy, but it also strongly restricts the class of functions for which the algorithm can be proven to be more efficient than other approaches.