Approximation of parametric derivatives by the empirical interpolation method (Q385450)

The authors develop a new a priori convergence theory for the approximation of parametric derivatives of parametrized functions. The best approximations to parametric derivatives in a sequence of approximation spaces generated by a general approximation scheme is considered and it is shown that these approximations are convergent provided that the best approximation to the function itself is convergent. Estimates for the convergence rates are also provided. The main limitations of the theory are related to regularity assumptions in space and parameter on the parametrized function, and on the particular norms that may be considered. It is claimed that the new theory also applies to cases in which the parametrized function is defined implicitly as the solution to a differential equation. In support of the new general theory two numerical results with spaces generated by a particular approximation scheme -- the empirical interpolation method -- are presented.