On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization

Atanu Mazumdar; Tinkle Chugh; Kaisa Miettinen; Manuel Lopez-Ibanez

doi:10.1007/978-3-030-12598-1_37

On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization

Atanu Mazumdar, Tinkle Chugh, Kaisa Miettinen, Manuel Lopez-Ibanez

Management Sciences

Research output: Chapter in Book/Conference proceeding › Conference contribution › peer-review

121 Downloads (Pure)

Abstract

Many works on surrogate-assisted evolutionary multiobjective optimization have been devoted to problems where function evaluations are time-consuming (e.g., based on simulations). In many real-life optimization problems, mathematical or simulation models are not always available and, instead, we only have data from experiments, measurements or sensors. In such cases, optimization is to be performed on surrogate models built on the data available. The main challenge there is to fit an accurate surrogate model and to obtain meaningful solutions.
We apply Kriging as a surrogate model and utilize corresponding uncertainty
information in different ways during the optimization process. We discuss experimental results obtained on benchmark multiobjective optimization problems with different sampling techniques and numbers of objectives. The results show the effect of different ways of utilizing uncertainty information on the quality of solutions.

Original language	English
Title of host publication	Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings
Editors	Sanaz Mostaghim, Carlos A. Coello Coello, Kathrin Klamroth, Patrick Reed, Kalyanmoy Deb, Erik Goodman, Kaisa Miettinen
Pages	463-474
Number of pages	12
DOIs	https://doi.org/10.1007/978-3-030-12598-1_37
Publication status	Published - 2019
Event	10th International Conference on Evolutionary Multi-Criterion Optimization - East Lansing, United States Duration: 10 Mar 2019 → 13 Mar 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11411 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	10th International Conference on Evolutionary Multi-Criterion Optimization
Abbreviated title	EMO19
Country/Territory	United States
City	East Lansing
Period	10/03/19 → 13/03/19

Keywords

Gaussian process
Machine learning
Metamodelling
Pareto optimality
Surrogate

Access to Document

10.1007/978-3-030-12598-1_37

samplepaperAccepted author manuscript, 657 KB

Cite this

Mazumdar, A., Chugh, T., Miettinen, K., & Lopez-Ibanez, M. (2019). On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization. In S. Mostaghim, C. A. Coello Coello, K. Klamroth, P. Reed, K. Deb, E. Goodman, & K. Miettinen (Eds.), Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings (pp. 463-474). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11411 LNCS). https://doi.org/10.1007/978-3-030-12598-1_37

Mazumdar, Atanu ; Chugh, Tinkle ; Miettinen, Kaisa et al. / On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization. Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings. editor / Sanaz Mostaghim ; Carlos A. Coello Coello ; Kathrin Klamroth ; Patrick Reed ; Kalyanmoy Deb ; Erik Goodman ; Kaisa Miettinen. 2019. pp. 463-474 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{bc27646c1c174e0583f9b6b95e72c829,

title = "On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization",

abstract = "Many works on surrogate-assisted evolutionary multiobjective optimization have been devoted to problems where function evaluations are time-consuming (e.g., based on simulations). In many real-life optimization problems, mathematical or simulation models are not always available and, instead, we only have data from experiments, measurements or sensors. In such cases, optimization is to be performed on surrogate models built on the data available. The main challenge there is to fit an accurate surrogate model and to obtain meaningful solutions.We apply Kriging as a surrogate model and utilize corresponding uncertaintyinformation in different ways during the optimization process. We discuss experimental results obtained on benchmark multiobjective optimization problems with different sampling techniques and numbers of objectives. The results show the effect of different ways of utilizing uncertainty information on the quality of solutions.",

keywords = "Gaussian process, Machine learning, Metamodelling, Pareto optimality, Surrogate",

author = "Atanu Mazumdar and Tinkle Chugh and Kaisa Miettinen and Manuel Lopez-Ibanez",

year = "2019",

doi = "10.1007/978-3-030-12598-1_37",

language = "English",

isbn = "9783030125974",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "463--474",

editor = "Sanaz Mostaghim and {Coello Coello}, {Carlos A.} and Kathrin Klamroth and Patrick Reed and Kalyanmoy Deb and Erik Goodman and Kaisa Miettinen",

booktitle = "Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings",

note = "10th International Conference on Evolutionary Multi-Criterion Optimization, EMO19 ; Conference date: 10-03-2019 Through 13-03-2019",

}

Mazumdar, A, Chugh, T, Miettinen, K & Lopez-Ibanez, M 2019, On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization. in S Mostaghim, CA Coello Coello, K Klamroth, P Reed, K Deb, E Goodman & K Miettinen (eds), Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11411 LNCS, pp. 463-474, 10th International Conference on Evolutionary Multi-Criterion Optimization, East Lansing, Michigan, United States, 10/03/19. https://doi.org/10.1007/978-3-030-12598-1_37

On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization. / Mazumdar, Atanu; Chugh, Tinkle; Miettinen, Kaisa et al.
Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings. ed. / Sanaz Mostaghim; Carlos A. Coello Coello; Kathrin Klamroth; Patrick Reed; Kalyanmoy Deb; Erik Goodman; Kaisa Miettinen. 2019. p. 463-474 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11411 LNCS).

Research output: Chapter in Book/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization

AU - Mazumdar, Atanu

AU - Chugh, Tinkle

AU - Miettinen, Kaisa

AU - Lopez-Ibanez, Manuel

PY - 2019

Y1 - 2019

N2 - Many works on surrogate-assisted evolutionary multiobjective optimization have been devoted to problems where function evaluations are time-consuming (e.g., based on simulations). In many real-life optimization problems, mathematical or simulation models are not always available and, instead, we only have data from experiments, measurements or sensors. In such cases, optimization is to be performed on surrogate models built on the data available. The main challenge there is to fit an accurate surrogate model and to obtain meaningful solutions.We apply Kriging as a surrogate model and utilize corresponding uncertaintyinformation in different ways during the optimization process. We discuss experimental results obtained on benchmark multiobjective optimization problems with different sampling techniques and numbers of objectives. The results show the effect of different ways of utilizing uncertainty information on the quality of solutions.

AB - Many works on surrogate-assisted evolutionary multiobjective optimization have been devoted to problems where function evaluations are time-consuming (e.g., based on simulations). In many real-life optimization problems, mathematical or simulation models are not always available and, instead, we only have data from experiments, measurements or sensors. In such cases, optimization is to be performed on surrogate models built on the data available. The main challenge there is to fit an accurate surrogate model and to obtain meaningful solutions.We apply Kriging as a surrogate model and utilize corresponding uncertaintyinformation in different ways during the optimization process. We discuss experimental results obtained on benchmark multiobjective optimization problems with different sampling techniques and numbers of objectives. The results show the effect of different ways of utilizing uncertainty information on the quality of solutions.

KW - Gaussian process

KW - Machine learning

KW - Metamodelling

KW - Pareto optimality

KW - Surrogate

UR - http://www.scopus.com/inward/record.url?scp=85063027062&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-12598-1_37

DO - 10.1007/978-3-030-12598-1_37

M3 - Conference contribution

SN - 9783030125974

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 463

EP - 474

BT - Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings

A2 - Mostaghim, Sanaz

A2 - Coello Coello, Carlos A.

A2 - Klamroth, Kathrin

A2 - Reed, Patrick

A2 - Deb, Kalyanmoy

A2 - Goodman, Erik

A2 - Miettinen, Kaisa

T2 - 10th International Conference on Evolutionary Multi-Criterion Optimization

Y2 - 10 March 2019 through 13 March 2019

ER -

Mazumdar A, Chugh T, Miettinen K, Lopez-Ibanez M. On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization. In Mostaghim S, Coello Coello CA, Klamroth K, Reed P, Deb K, Goodman E, Miettinen K, editors, Evolutionary Multi-Criterion Optimization - 10th International Conference, EMO 2019, Proceedings. 2019. p. 463-474. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). Epub 2019 Feb 3. doi: 10.1007/978-3-030-12598-1_37

On Dealing with Uncertainties from Kriging Models in Offline Data-driven Evolutionary Multiobjective Optimization

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this