On sampling methods for costly multi-objective black-box optimization

Ingrida Steponavice, Mojdeh Shirazi Manesh, Rob Hyndman, Kate Smith-Miles, Laura Villanova

Research output: Chapter in Book/Report/Conference proceedingChapter (Book)Researchpeer-review

Abstract

We investigate the impact of different sampling techniques on the performance of multi-objective optimization methods applied to costly black-box optimization problems. Such problems are often solved using an algorithm in which a surrogate model approximates the true objective function and provides predicted objective values at a lower cost. As the surrogate model is based on evaluations of a small number of points, the quality of the initial sample can have a great effect on the overall effectiveness of the optimization. In this study, we demonstrate how various sampling techniques affect the results of applying different optimization algorithms to a set of benchmark problems. Additionally, some recommendations on usage of sampling methods are provided.
Original languageEnglish
Title of host publicationAdvances in Stochastic and Deterministic Global Optimization
EditorsPanos M. Pardalos, Anatoly Zhigljavsky, Julius Žilinskas
Place of PublicationSwitzerland
PublisherSpringer
Pages273-296
Number of pages24
Volume107
ISBN (Electronic)9783319299754
ISBN (Print)9783319299730
DOIs
Publication statusPublished - 5 Nov 2016

Publication series

NameSpringer Optimization and Its Applications
PublisherSpringer International Publishing
NumberPart III
Volume107

Keywords

  • Design of experiment
  • Space-filling
  • Low-discrepancy
  • Efficient global optimization

Cite this

Steponavice, I., Shirazi Manesh, M., Hyndman, R., Smith-Miles, K., & Villanova, L. (2016). On sampling methods for costly multi-objective black-box optimization. In P. M. Pardalos, A. Zhigljavsky, & J. Žilinskas (Eds.), Advances in Stochastic and Deterministic Global Optimization (Vol. 107, pp. 273-296). (Springer Optimization and Its Applications; Vol. 107, No. Part III). Switzerland: Springer. https://doi.org/10.1007/978-3-319-29975-4_15
Steponavice, Ingrida ; Shirazi Manesh, Mojdeh ; Hyndman, Rob ; Smith-Miles, Kate ; Villanova, Laura. / On sampling methods for costly multi-objective black-box optimization. Advances in Stochastic and Deterministic Global Optimization. editor / Panos M. Pardalos ; Anatoly Zhigljavsky ; Julius Žilinskas. Vol. 107 Switzerland : Springer, 2016. pp. 273-296 (Springer Optimization and Its Applications; Part III).
@inbook{ff573c8b3ac64b43b23f62874c93c212,
title = "On sampling methods for costly multi-objective black-box optimization",
abstract = "We investigate the impact of different sampling techniques on the performance of multi-objective optimization methods applied to costly black-box optimization problems. Such problems are often solved using an algorithm in which a surrogate model approximates the true objective function and provides predicted objective values at a lower cost. As the surrogate model is based on evaluations of a small number of points, the quality of the initial sample can have a great effect on the overall effectiveness of the optimization. In this study, we demonstrate how various sampling techniques affect the results of applying different optimization algorithms to a set of benchmark problems. Additionally, some recommendations on usage of sampling methods are provided.",
keywords = "Design of experiment, Space-filling, Low-discrepancy, Efficient global optimization",
author = "Ingrida Steponavice and {Shirazi Manesh}, Mojdeh and Rob Hyndman and Kate Smith-Miles and Laura Villanova",
year = "2016",
month = "11",
day = "5",
doi = "10.1007/978-3-319-29975-4_15",
language = "English",
isbn = "9783319299730",
volume = "107",
series = "Springer Optimization and Its Applications",
publisher = "Springer",
number = "Part III",
pages = "273--296",
editor = "Pardalos, {Panos M. } and Zhigljavsky, {Anatoly } and Žilinskas, {Julius }",
booktitle = "Advances in Stochastic and Deterministic Global Optimization",

}

Steponavice, I, Shirazi Manesh, M, Hyndman, R, Smith-Miles, K & Villanova, L 2016, On sampling methods for costly multi-objective black-box optimization. in PM Pardalos, A Zhigljavsky & J Žilinskas (eds), Advances in Stochastic and Deterministic Global Optimization. vol. 107, Springer Optimization and Its Applications, no. Part III, vol. 107, Springer, Switzerland, pp. 273-296. https://doi.org/10.1007/978-3-319-29975-4_15

On sampling methods for costly multi-objective black-box optimization. / Steponavice, Ingrida; Shirazi Manesh, Mojdeh; Hyndman, Rob; Smith-Miles, Kate; Villanova, Laura.

Advances in Stochastic and Deterministic Global Optimization. ed. / Panos M. Pardalos; Anatoly Zhigljavsky; Julius Žilinskas. Vol. 107 Switzerland : Springer, 2016. p. 273-296 (Springer Optimization and Its Applications; Vol. 107, No. Part III).

Research output: Chapter in Book/Report/Conference proceedingChapter (Book)Researchpeer-review

TY - CHAP

T1 - On sampling methods for costly multi-objective black-box optimization

AU - Steponavice, Ingrida

AU - Shirazi Manesh, Mojdeh

AU - Hyndman, Rob

AU - Smith-Miles, Kate

AU - Villanova, Laura

PY - 2016/11/5

Y1 - 2016/11/5

N2 - We investigate the impact of different sampling techniques on the performance of multi-objective optimization methods applied to costly black-box optimization problems. Such problems are often solved using an algorithm in which a surrogate model approximates the true objective function and provides predicted objective values at a lower cost. As the surrogate model is based on evaluations of a small number of points, the quality of the initial sample can have a great effect on the overall effectiveness of the optimization. In this study, we demonstrate how various sampling techniques affect the results of applying different optimization algorithms to a set of benchmark problems. Additionally, some recommendations on usage of sampling methods are provided.

AB - We investigate the impact of different sampling techniques on the performance of multi-objective optimization methods applied to costly black-box optimization problems. Such problems are often solved using an algorithm in which a surrogate model approximates the true objective function and provides predicted objective values at a lower cost. As the surrogate model is based on evaluations of a small number of points, the quality of the initial sample can have a great effect on the overall effectiveness of the optimization. In this study, we demonstrate how various sampling techniques affect the results of applying different optimization algorithms to a set of benchmark problems. Additionally, some recommendations on usage of sampling methods are provided.

KW - Design of experiment

KW - Space-filling

KW - Low-discrepancy

KW - Efficient global optimization

U2 - 10.1007/978-3-319-29975-4_15

DO - 10.1007/978-3-319-29975-4_15

M3 - Chapter (Book)

SN - 9783319299730

VL - 107

T3 - Springer Optimization and Its Applications

SP - 273

EP - 296

BT - Advances in Stochastic and Deterministic Global Optimization

A2 - Pardalos, Panos M.

A2 - Zhigljavsky, Anatoly

A2 - Žilinskas, Julius

PB - Springer

CY - Switzerland

ER -

Steponavice I, Shirazi Manesh M, Hyndman R, Smith-Miles K, Villanova L. On sampling methods for costly multi-objective black-box optimization. In Pardalos PM, Zhigljavsky A, Žilinskas J, editors, Advances in Stochastic and Deterministic Global Optimization. Vol. 107. Switzerland: Springer. 2016. p. 273-296. (Springer Optimization and Its Applications; Part III). https://doi.org/10.1007/978-3-319-29975-4_15