HR: 10:35h
AN: NG42A-02 [Abstracts]
TI: Automatic Dfferentiation in Geophysical Inverse Problems
AU: * Sambridge, M
EM: malcolm.sambridge@anu.edu.au
AF: Research School of Earth Sciences, Australian National University, Canberra, ACT 0200
AU: Rickwood, P
EM: Peter.Rickwood@anu.edu.au
AF: Research School of Earth Sciences, Australian National University, Canberra, ACT 0200
AU: Rawlinson, N
EM: nick@rses.anu.edu.au
AF: Research School of Earth Sciences, Australian National University, Canberra, ACT 0200
AB:
Automatic differentiation (AD) is the technique whereby `output variables' of a computer code evaluating some complicated
function or solving some differential equation can be differentiated with respect to `input variables', without the need for
deriving and coding up of explicit mathematical formulae. The great promise of AD is that it combines the generality of
finite difference techniques and the accuracy and efficiency of analytical derivatives, while at the same time eliminating
`human' coding errors.
Over the next ten years Automatic Differentiation is set to have a major impact in the area of nonlinear optimization.
Similarly AD tools have considerable potential for use in linearizing nonlinear inverse problems in geophysics and also for
sensitivity analysis of large numerical calculations, e.g. wave propagation and mantle convection. At present, however, AD
tools appear to be little used
in the geosciences.
Here we present some results of using a state of the art AD tool to perform source to source code translation in a range of
geoscience problems. These include calculating derivatives for Gibbs Free energy minimization, seismic receiver functions,
ray tracing and wavefront tracking codes. Issues of accuracy and efficiency will be discussed and examples of success and
failure presented.
DE: 0500 COMPUTATIONAL GEOPHYSICS (3200, 3252, 7833)
DE: 3200 MATHEMATICAL GEOPHYSICS (0500, 4400, 7833)
DE: 3260 Inverse theory
DE: 7290 Computational seismology
SC: Nonlinear Geophysics [NG]
MN: Fall Meeting 2005