The need for knowledge and reliability in numeric computation: Case study of multivariate Padé approximation (Q1327438)

The authors review results of their theoretical and numerical investigations on the multivariate Padé approximation. They show how the knowledge about the problem environment can guide a number of choices needed before the computation. Secondly, they show how the defect correction algorithm adapted for interval arithmetic allows to improve numerical results. The computation of bivariate Padé approximation is illustrated in the case of the beta function.