Engineering Mechanics

Proceedings Vol. 22 (2016)

ENGINEERING MECHANICS 2016

Copyright © 2016 Institute of Thermomechanics, Academy of Sciences of the Czech Republic, v.v.i., Prague

 

Myšáková E., Lepš M.
pages 441 - 444, full text

Surrogate modeling (Meta-modeling) is an often used tool for analysis of behavior of complex systems which are usually described by computationally demanding models. Surrogate models provide an approximation of the original model's response in a fraction of time and therefore are suitable when multiple evaluations are needed. Many types of meta-models exist and each suits another type of problem. On the other hand it is not always possible to select the right meta-model in advance. Therefore parallel construction of several meta-models and their subsequent comparison and combining can be utilized with advantage. A typical method called PRESS weighted average surrogate which uses the prediction sum of squares obtained by cross-validation for computation of the weights for linear combination of individual surrogates is discussed in this contribution and illustrated on several 2-dimensional benchmark examples using a group of different meta-models.

Text and facts may be copied and used freely, but credit should be given to these Proceedings.

All papers were reviewed by members of the scientific committee.