Surrogate modeling with sequential design for design and analysis of electronic systems

The growing computational demands of modern engineering simulations as used frequently in fields ranging from computational fluid dynamics to electromagnetics, requires methodologies to be able to perform evaluation intensive tasks. Popular analyses include design space exploration, visualization, o...

Full description

Saved in:
Bibliographic Details
Published in2017 International Conference on Electromagnetics in Advanced Applications (ICEAA) pp. 1403 - 1406
Main Authors van der Herten, J., Dutordoir, V., Couckuyt, I., Dhaene, T.
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.09.2017
Subjects
Adaptation models
Computational modeling
Data models
Gaussian processes
Mathematical model
Object oriented modeling
Optimization
Online AccessGet full text

Cover

More Information
Summary:The growing computational demands of modern engineering simulations as used frequently in fields ranging from computational fluid dynamics to electromagnetics, requires methodologies to be able to perform evaluation intensive tasks. Popular analyses include design space exploration, visualization, optimization or sensitivity analysis. This work provides an overview of advancements in surrogate modeling, a data-driven approximation technique. Both sequential design and adaptive modeling are covered, and an integrated platform for surrogate modeling is presented. Finally, a recent technique known as deep Gaussian processes is highlighted as a promising alternative for surrogate modeling of non-stationary response surfaces.
DOI:10.1109/ICEAA.2017.8065540