Multistep approximation algorithms: Improved convergence rates through postconditioning with smoothing kernels (Q1284137)

Certain widely used multistep approximation algorithms are interpreted as instances of an approximate Newton method. In an earlier paper [\textit{J. W. Jerome}, Numer. Math. 47, 123-138 (1985; Zbl 0579.65046)] the second author showed that the convergence rates of approximate Newton methods suffer from ``loss of derivative'', and that the subsequent linear rate of convergence can be improved to be superlinear using an adaptation of Nash-Moser iteration for numerical analysis purposes. The essence of this adaptation is a splitting of the inversion and the smoothing into two separate steps. In the present paper the authors apply these ideas to scattered data approximation as well as the numerical solution of partial differential equations. Several radial kernels for the smoothing operation are investigated . Results of Hörmander and the second author are generalized to Sobolev and Besov spaces. The resulting theory provides convergence results for certain multilevel approximation algorithms. The established convergence rate is superlinear if a smoothing operation is included in the algorithm as a postconditioner. Numerical examples conclude the paper suggesting the power of the theory.