An Experimental Algorithm for N-Dimensional Adaptive Quadrature

A working experimental program for n-dimensional adaptive quadrature is discussed. The program uses the simplex rather than the cube as the basic ''cell.'' Symmetric interpolatory quadratures of variable order are derived during execution and used as basic integrators. Integrand values are stored into a table by means of a double hashing technique for efficient retrieval. Iterated calls are allowed with the dimension of the region factored arbitrarily. Restarting the calculation for greater precision costs no more than if the user had requested the higher accuracy in the first place. 2 figures.