|
Organizers |
Computational Differentiation Applied to Neutron Scattering
by
H. Martin Bücker
Institute for Scientific Computing, Aachen University of Technology
Coauthors: Dieter an Mey (Computing Center, Aachen University of Technology)
Quasielastic neutron scattering is a method to investigate the dynamics of molecular liquids. At the Institute of Physical Chemistry, Technical University of Aachen, neutron scattering experiments are carried out and their results are compared to a parametrized theoretical model. The parameters are determined by fitting experimental and theoretical results using a nonlinear least-squares problem. Automatic differentiation is used to switch from a quasi-Newton method to a Gauss-Newton method in the underlying simulation code. Furthermore, rather than mechanically applying techniques of automatic differentiation in the traditional sense, it is shown that the combination of automatic differentiation and high-level mathematical knowledge can be used to improve performance significantly. More precisely, a single subroutine of the Adifor-generated differentiated code is exchanged by a few lines of hand-written code based on the Cauchy-Riemann equations.
http://www.sc.rwth-aachen.de/buecker/ad2000.html
Date received: February 11, 2000
Copyright © 2000 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 # cads-71.