Continuous simulation optimization with model mismatch using Gaussian process regression

Alireza Inanlouganji, Giulia Pedrielli, Georgios Fainekos, Sebastian Pokutta

Research output: Chapter in Book/Report/Conference proceedingConference contribution

5 Scopus citations


Multi-fidelity simulation optimization is an emerging area looking at the use of low-fidelity (computationally cheap but inaccurate) models to optimize high-fidelity (expensive and accurate) models. In this context, low-fidelity models exhibit a mismatch to high-fidelity models whose values can be point-wise obtained by querying an expensive simulator. Herein, an efficient multi-fidelity algorithm is proposed for continuous global optimization. The algorithm is made up of an additive model that consolidates low-fidelity and bias (mismatch) predictions. Two sampling criteria with different use of the cumulated high and low-fidelity information are introduced as well as a cheap certificate guiding the decision on whether to sample from the expensive simulator. The performance of proposed algorithms is evaluated using a state of the art stochastic search benchmark algorithm. The results show that the proposed methods can beat the benchmark with improved accuracy, while essentially maintaining the same performance in terms of number of expensive simulations.

Original languageEnglish (US)
Title of host publicationWSC 2018 - 2018 Winter Simulation Conference
Subtitle of host publicationSimulation for a Noble Cause
PublisherInstitute of Electrical and Electronics Engineers Inc.
Number of pages12
ISBN (Electronic)9781538665725
StatePublished - Jan 31 2019
Event2018 Winter Simulation Conference, WSC 2018 - Gothenburg, Sweden
Duration: Dec 9 2018Dec 12 2018

Publication series

NameProceedings - Winter Simulation Conference
ISSN (Print)0891-7736


Conference2018 Winter Simulation Conference, WSC 2018

ASJC Scopus subject areas

  • Software
  • Modeling and Simulation
  • Computer Science Applications


Dive into the research topics of 'Continuous simulation optimization with model mismatch using Gaussian process regression'. Together they form a unique fingerprint.

Cite this