When working with numerical models, it is essential to determine model parameters which are as realistic as possible. Optimization techniques are used more and more frequently to solve this task. However, using these methods may lead to very high time costs – in particular, if rather complicated forward calculations are involved. In this paper, we present a class of methods wich allows estimating the solution of this kind of optimization problems, based on relatively few sampling points. We put very weak constraints on the sampling point distribution; hence, they may be taken from previous forward calculations as well as from alternative sources. Starting from an introduction into the theoretical approach, a strategy for speeding up inverse optimization problems is introduced which is illustrated by an example from geomechanics.
原文链接
分享
引用
