|
Organizers |
A preconditioned GMRES method for fast fitting of kriging surfaces
by
Cameron Mouat
University of Canterbury
Coauthors: Rick Beatson
Kriging is a surface fitting method that has been often used by researchers in mining engineering, soil science, hydrology and atmospheric science. Evaluating the surface at a point is frequently done by using only a subset of the data and then inverting a small system for each point. This technique however gives a surface which is not continuous and clearly not all information is used at each evaluation point. In the past to fit the surface using all the data has required O(N3) operations and O(N2) storage, where N is the number of data points. This paper considers computing the radial basis function like parameters of a Kriged surface with a combination of a fast matrix-vector multiplication for general \phi and the preconditioned GMRES iterative method. This combination reduces the computational cost of fitting to O(N logN) operations and O(N) storage. Evaluation of the fitted surface at a point then requires only O(1) operations.
Date received: August 9, 1999
Copyright © 1999 by the author(s). The author(s) of this document and the organizers of the conference have granted their consent to include this abstract in Atlas Conferences Inc. Document # cadr-10.