Automatic Differentiation Tools in Optimization Software
by
Jorge More
ANL

Despite advances in automatic differentiation algorithms and software, there is disagreement on the value of incorporating automatic differentiation tools in optimization software. There are various reasons for this state of affairs. An important reason seems to be that there is little published experience on the effect of automatic differentiation tools on realistic problems, and thus users worry that automatic differentiations tools are not applicable to their problems or too expensive in terms of time or memory. Whatever the reasons, it is certainly the case that few optimization codes incorporate automatic differentiation tools.

Our aim is to develop optimization software for large problems that only requires the specification of the problem. All other quantities required by the software (gradients, Jacobians, sparsity patterns, ...) should be generated automatically. In this talk we review current experience in the use of automatic differentiation tools with the emphasis on software for large-scale optimization problems and systems of nonlinear equations.

Date received: April 17, 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 # caev-09.