© 2001 by Institute of Mathematics and its Applications
Local polynomial reproduction and moving least squares approximation
1 Institut für Numerische und Angewandte Mathematik, Universität Göttingen, Lotzestraße 16-18, D-37083 Göttingen, Germany
Local polynomial reproduction is a key ingredient in providing error estimates for several approximation methods. To bound the Lebesgue constants is a hard task especially in a multivariate setting. We provide a result which allows us to bound the Lebesgue constants uniformly and independently of the space dimension by oversampling. We get explicit and small bounds for the Lebesgue constants. Moreover, we use these results to establish error estimates for the moving least squares approximation scheme, also with special emphasis on the involved constants. We discuss the numerical treatment of the method and analyse its effort. Finally, we give large scale examples.
Key Words: scattered data approximation, approximation orders
Received 6 October 1999. Accepted 8 March 2000.