Seminar

The advent of reliable computing machines, computer algebra systems, and multiple precision computational packages diminished the need for tables of reference values for computing function values by interpolation, but today's numerical analysts, scientific researchers, and software developers still need a way to confirm the accuracy of numerical algorithms that compute mathematical function values. The field of validated computation of mathematical functions explores the development of multiple precision codes that compute certifiably accurate function values that can be used to test the accuracy of function data from personal, commercial, or publicly available codes. We discuss the analysis used to obtain reliable error bounds for floating point approximations and describe the implementation of the work in a publicly available beta site.