Exploration of Metamodeling Sampling Criteria for Constrained Global Optimization

By M. Sasena, Panos Y. Papalambros, Pierre Goovaerts. 2002

465 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Statistical and Nonlinear Physics

View Paper: https://www.semanticscholar.org/paper/8fb539890f64b86fd1fc586054849149fdeabdf6

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Statistical and Nonlinear Physics

Citations in the corpus

Kriging is well-suited to parallelize optimization (2010, unknown): “There is a detailed review of optimization methods relying on a metamodel in [44, 45] or [20].”
Multiobjective Optimization on a Limited Budget of Evaluations Using Model-Assisted -Metric Selection (2008, background): “However, approaches have been developed to overcome this drawback [14,15].”
On the Design of Optimization Strategies Based on Global Response Surface Approximation Models (2005, background): “Sasena et al. (2002) suggest a heuristic reminiscent of Simulated Annealing, which is based on generalized expected improvement.”
Global Optimization of Stochastic Black-Box Systems via Sequential Kriging Meta-Models (2006, unknown): “Keywords: Efficient Global Optimization, Stochastic Black-box Systems, Expected Improvement, Kriging”
Provably Convergent Multifidelity Optimization Algorithm Not Requiring High-Fidelity Derivatives (2012, unknown): “fidelity function value [19].”
Infill sampling criteria for surrogate-based optimization with constraint handling (2012, background): “To name a few: generalized expected improvement (Schonlau 1997, Sasena et al. 2002) and weighted expected improvement (Sóbester et al. 2005) assign different weightings to exploration and exploitation; quantile-based expected improvement (Picheny et al. 2010) deals with noisy experiments; and…”
Application-driven sequential designs for simulation experiments: Kriging metamodelling (2004, methodology): “In the literature (including Sasena et al. 2002 and Jones et al. 1998), we did not find an appealing stopping criterion for our sequential design; future research may be needed.”
An informational approach to the global optimization of expensive-to-evaluate functions (2006, methodology): “Numerous illustrations of this superiority can be found in the literature (see, for instance, [6]) and many variations have been explored (for extensive surveys, see [13] and [17]).”
Cooling Strategies for the Moment-Generating Function in Bayesian Global Optimization (2018, background): “[18] propose to build a cooling table, where g is initialized wit high value, making the initial search explorative and then g is gradually decreased to enable the fast convergence.”
Reduced-Order Nonlinear Unsteady Aerodynamic Modeling Using a Surrogate-Based Recurrence Framework (2010, methodology): “by using more “intelligent” sampling methods [34,42] compared to purely space-filling approaches.”
A Provably Convergent Multifidelity Optimization Algorithm not Requiring High-Fidelity Derivatives (2010, methodology): “For the EGO implementation, 4n high-fidelity evaluations are used for the initial Latin hypercube sample and a genetic algorithm is used to maximize the expected improvement.”
Constrained Efficient Global Optimization of Expensive Black-box Functions (2022, methodology): “…is where constraint violations are allowed to occur during the optimization process without penalty, for which there exists generic constrained Bayesian (efficient global) optimization methods (Schonlau, 1997; Sasena et al., 2002; Basudhar et al., 2012; Bagheri et al., 2017; Zhang et al., 2021).”
Sequential kriging optimization using multiple-fidelity evaluations (2006, methodology): “When constraints are inexpensive, we can simply maximize the infill sampling criteria as a constrained optimization problem, as proposed by Sasena et al. (2002). However, when constraints are expensive, we may need to utilize metamodels; thus, the problem is complicated due to the prediction uncerta”