Simulation Optimization via Bootstrapped Kriging

Tutorial (Replaced by CentER DP 2013-064)

Research output: Working paperDiscussion paperOther research output

Abstract

Kriging (or Gaussian Process) metamodels may be analyzed through bootstrapping, which is a versatile statistical method but must be adapted to the speci c problem being analyzed. More precisely, a random or discrete-event simulation may be run several times for the same scenario (combination of simulation inputs); the resulting replicated responses may be resampled with replacement, which is called "distribution-free bootstrapping". In engineering, however, deterministic simulation is often applied; such a simulation is run only once for the same scenario, so "parametric bootstrapping" is used. This bootstrapping assumes a multivariate Gaussian distribution, which is sampled after its parameters are estimated from the simulation's input/output data. More specifically, this tutorial covers the following recent approaches: (1) E¢ cient Global Optimization (EGO) via Expected Improvement (EI) using parametric bootstrapping to obtain an estimator of the Kriging predictor's variance accounting for the randomness resulting from estimating the Kriging parameters. (2) Constrained optimization via Mathematical Programming applied to Kriging metamodels using distribution-free bootstrapping to validate these metamodels. (3) Robust optimization accounting for an environment that is not exactly known (so it is uncertain); this optimization may use Mathematical Programming and Kriging with distribution-free bootstrapping to estimate the Pareto frontier. (4) Assuming a characteristic like monotonicity for the outputs as a function of the inputs, bootstrapped Kriging may preserve this characteristic.
Original languageEnglish
Place of PublicationTilburg
PublisherInformation Management
Volume2011-064
Publication statusPublished - 2011

Publication series

NameCentER Discussion Paper
Volume2011-064

Fingerprint

Mathematical programming
Constrained optimization
Gaussian distribution
Discrete event simulation
Global optimization
Statistical methods

Keywords

  • simulation
  • optimization
  • experimental design
  • stochastic processes
  • engineering

Cite this

Kleijnen, J. P. C. (2011). Simulation Optimization via Bootstrapped Kriging: Tutorial (Replaced by CentER DP 2013-064). (CentER Discussion Paper; Vol. 2011-064). Tilburg: Information Management.
@techreport{3fa12254b6fa4408ae7d851446148161,
title = "Simulation Optimization via Bootstrapped Kriging: Tutorial (Replaced by CentER DP 2013-064)",
abstract = "Kriging (or Gaussian Process) metamodels may be analyzed through bootstrapping, which is a versatile statistical method but must be adapted to the speci c problem being analyzed. More precisely, a random or discrete-event simulation may be run several times for the same scenario (combination of simulation inputs); the resulting replicated responses may be resampled with replacement, which is called {"}distribution-free bootstrapping{"}. In engineering, however, deterministic simulation is often applied; such a simulation is run only once for the same scenario, so {"}parametric bootstrapping{"} is used. This bootstrapping assumes a multivariate Gaussian distribution, which is sampled after its parameters are estimated from the simulation's input/output data. More specifically, this tutorial covers the following recent approaches: (1) E¢ cient Global Optimization (EGO) via Expected Improvement (EI) using parametric bootstrapping to obtain an estimator of the Kriging predictor's variance accounting for the randomness resulting from estimating the Kriging parameters. (2) Constrained optimization via Mathematical Programming applied to Kriging metamodels using distribution-free bootstrapping to validate these metamodels. (3) Robust optimization accounting for an environment that is not exactly known (so it is uncertain); this optimization may use Mathematical Programming and Kriging with distribution-free bootstrapping to estimate the Pareto frontier. (4) Assuming a characteristic like monotonicity for the outputs as a function of the inputs, bootstrapped Kriging may preserve this characteristic.",
keywords = "simulation, optimization, experimental design, stochastic processes, engineering",
author = "Kleijnen, {Jack P.C.}",
year = "2011",
language = "English",
volume = "2011-064",
series = "CentER Discussion Paper",
publisher = "Information Management",
type = "WorkingPaper",
institution = "Information Management",

}

Kleijnen, JPC 2011 'Simulation Optimization via Bootstrapped Kriging: Tutorial (Replaced by CentER DP 2013-064)' CentER Discussion Paper, vol. 2011-064, Information Management, Tilburg.

Simulation Optimization via Bootstrapped Kriging : Tutorial (Replaced by CentER DP 2013-064). / Kleijnen, Jack P.C.

Tilburg : Information Management, 2011. (CentER Discussion Paper; Vol. 2011-064).

Research output: Working paperDiscussion paperOther research output

TY - UNPB

T1 - Simulation Optimization via Bootstrapped Kriging

T2 - Tutorial (Replaced by CentER DP 2013-064)

AU - Kleijnen, Jack P.C.

PY - 2011

Y1 - 2011

N2 - Kriging (or Gaussian Process) metamodels may be analyzed through bootstrapping, which is a versatile statistical method but must be adapted to the speci c problem being analyzed. More precisely, a random or discrete-event simulation may be run several times for the same scenario (combination of simulation inputs); the resulting replicated responses may be resampled with replacement, which is called "distribution-free bootstrapping". In engineering, however, deterministic simulation is often applied; such a simulation is run only once for the same scenario, so "parametric bootstrapping" is used. This bootstrapping assumes a multivariate Gaussian distribution, which is sampled after its parameters are estimated from the simulation's input/output data. More specifically, this tutorial covers the following recent approaches: (1) E¢ cient Global Optimization (EGO) via Expected Improvement (EI) using parametric bootstrapping to obtain an estimator of the Kriging predictor's variance accounting for the randomness resulting from estimating the Kriging parameters. (2) Constrained optimization via Mathematical Programming applied to Kriging metamodels using distribution-free bootstrapping to validate these metamodels. (3) Robust optimization accounting for an environment that is not exactly known (so it is uncertain); this optimization may use Mathematical Programming and Kriging with distribution-free bootstrapping to estimate the Pareto frontier. (4) Assuming a characteristic like monotonicity for the outputs as a function of the inputs, bootstrapped Kriging may preserve this characteristic.

AB - Kriging (or Gaussian Process) metamodels may be analyzed through bootstrapping, which is a versatile statistical method but must be adapted to the speci c problem being analyzed. More precisely, a random or discrete-event simulation may be run several times for the same scenario (combination of simulation inputs); the resulting replicated responses may be resampled with replacement, which is called "distribution-free bootstrapping". In engineering, however, deterministic simulation is often applied; such a simulation is run only once for the same scenario, so "parametric bootstrapping" is used. This bootstrapping assumes a multivariate Gaussian distribution, which is sampled after its parameters are estimated from the simulation's input/output data. More specifically, this tutorial covers the following recent approaches: (1) E¢ cient Global Optimization (EGO) via Expected Improvement (EI) using parametric bootstrapping to obtain an estimator of the Kriging predictor's variance accounting for the randomness resulting from estimating the Kriging parameters. (2) Constrained optimization via Mathematical Programming applied to Kriging metamodels using distribution-free bootstrapping to validate these metamodels. (3) Robust optimization accounting for an environment that is not exactly known (so it is uncertain); this optimization may use Mathematical Programming and Kriging with distribution-free bootstrapping to estimate the Pareto frontier. (4) Assuming a characteristic like monotonicity for the outputs as a function of the inputs, bootstrapped Kriging may preserve this characteristic.

KW - simulation

KW - optimization

KW - experimental design

KW - stochastic processes

KW - engineering

M3 - Discussion paper

VL - 2011-064

T3 - CentER Discussion Paper

BT - Simulation Optimization via Bootstrapped Kriging

PB - Information Management

CY - Tilburg

ER -

Kleijnen JPC. Simulation Optimization via Bootstrapped Kriging: Tutorial (Replaced by CentER DP 2013-064). Tilburg: Information Management. 2011. (CentER Discussion Paper).