Victor Pereyra
Weidlinger Associates
Los Altos, CA

“A Toolkit for Optimization Involving Large Scale Simulations”

Abstract

The last part of the 20th Century has seen an explosion in the complexity and size of engineering and scientific simulations of natural and mechanical phenomena that have been attempted successfully. This progress has been driven by a combination of better Applied and Computational Mathematics and by Moore's law of computing.

Although not all the simulation problems are yet solved, it has become clear that simulation is not usually an end in itself, but rather the final goal involves an optimization process. We see this clearly in problems of Optimal Design, Inverse problems in many disciplines, material identification, and in general in the fitting of models to experimental data.

In this talk we will dwell on some aspects of this fascinating new/old field and discuss in particular a toolkit of general techniques that we are building in order to be able to deploy them to solve new problems with a minimum of recoding.