Uncertainty Management Using Sequential Parameter Optimization

Bartz-Beielstein, Thomas; Jung, Christian; Zaefferer, Martin

doi:10.1007/978-1-4899-7547-8_4

Cited by 4 publications

(2 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The space-filling approach can treat all the regions of design space equally. LHS has been commonly defined for designing sample points based on the space-filling concept [66,67]. Undoubtedly, the simplest and the most popular sampling design methods utilize LHS [17,18].…”

Section: Step 1: Design Training Sample Pointsmentioning

confidence: 99%

A New Approach for Low-Dimensional Constrained Engineering Design Optimization Using Design and Analysis of Simulation Experiments

Parnianifard¹,

Chancharoen²,

Phanomchoeng³

et al. 2020

IJCIS

View full text Add to dashboard Cite

The number of function evaluations in many industrial applications of simulation-based optimization problems is strictly limited. Therefore, only little analytical information on objective and constraint functions is available. This paper presents an adaptive algorithm called the Surrogate-Based Constrained Global-Optimization (SCGO) method to solve black-box constrained simulation-based optimization problems involving computationally expensive objective function and inequality constraints. Firstly, Kriging surrogate is constructed over a new overall objective function (called loss function) to approximate the behavior of a true model. Then, an adaptive approach is provided to improve the optimal results sequentially while enforcing a feasible solution. The SCGO method is tested on several classical engineering design problems namely design of a tension/compression spring, design of a welded beam, design of a pressure vessel, and three-bar truss design. The results demonstrate that SCGO has advantages in solving the costly constrained problems and needs less costly function evaluations. Optimization results prove that the proposed algorithm is very competitive compared to the state-of-the-art metaheuristic algorithms.

show abstract

Section: Step 1: Design Training Sample Pointsmentioning

confidence: 99%

A New Approach for Low-Dimensional Constrained Engineering Design Optimization Using Design and Analysis of Simulation Experiments

Parnianifard¹,

Chancharoen²,

Phanomchoeng³

et al. 2020

IJCIS

View full text Add to dashboard Cite

show abstract

“…LHS has been commonly defined for designing computer experiments based on the space-filling concept. 39,40 Undoubtedly, the simplest and most popular computer experiment designs use LHS. 3,4 In general, for n input variables, m sample points are produced randomly into m intervals or scenarios (with equal probability).…”

Section: Metamodel-based Robust Simulation–optimizationmentioning

confidence: 99%

Comparative study of metamodeling and sampling design for expensive and semi-expensive simulation models under uncertainty

et al. 2019

View full text Add to dashboard Cite

In spite of the wide improvements in computer simulation packages, many complex simulation models, particularly under uncertainty, may be inefficient to run in terms of time, computation, and resources. To address such a challenge, integrating metamodels and robust design optimization has been applied. In the current paper, a systematic comparative study is implemented to evaluate the performance of three common metamodels, namely polynomial regression, kriging, and radial basis function. The required experiments are designed by different space-filling methods including the orthogonal array design and three forms of Latin hypercube sampling such as randomized, maximin, and correlation approaches. Although, the impact of sample size on the performance of metamodels in robust optimization results are investigated. All methods are analyzed using five two-dimensional test problems and one engineering problem while all of them are considered in two forms that are expensive (with a small sample size) and semi-expensive (with a large sample size). Uncertainty is assumed in all problems as a source of variability, so all test problems are conducted in the format of robust optimization in the class of dual response surface in order to estimate robust Pareto frontier. The performances of methods are studied in two terms of accuracy and robustness. Finally, the results of comparison, an applicable guideline is provided to aid the practitioners in selecting the appropriate combination of metamodels and sampling design methods for investigating set of robust optimal points (estimated Pareto frontier) in simulation–optimization problems under uncertainty.

show abstract

Sparsifying to optimize over multiple information sources: an augmented Gaussian process based algorithm

Candelieri

Archetti

2021

Struct Multidisc Optim

View full text Add to dashboard Cite

Optimizing a black-box, expensive, and multi-extremal function, given multiple approximations, is a challenging task known as multi-information source optimization (MISO), where each source has a different cost and the level of approximation (aka fidelity) of each source can change over the search space. While most of the current approaches fuse the Gaussian processes (GPs) modelling each source, we propose to use GP sparsification to select only “reliable” function evaluations performed over all the sources. These selected evaluations are used to create an augmented Gaussian process (AGP), whose name is implied by the fact that the evaluations on the most expensive source are augmented with the reliable evaluations over less expensive sources. A new acquisition function, based on confidence bound, is also proposed, including both cost of the next source to query and the location-dependent approximation of that source. This approximation is estimated through a model discrepancy measure and the prediction uncertainty of the GPs. MISO-AGP and the MISO-fused GP counterpart are compared on two test problems and hyperparameter optimization of a machine learning classifier on a large dataset.

show abstract

Uncertainty Management Using Sequential Parameter Optimization

Cited by 4 publications

References 20 publications

A New Approach for Low-Dimensional Constrained Engineering Design Optimization Using Design and Analysis of Simulation Experiments

A New Approach for Low-Dimensional Constrained Engineering Design Optimization Using Design and Analysis of Simulation Experiments

Comparative study of metamodeling and sampling design for expensive and semi-expensive simulation models under uncertainty

Sparsifying to optimize over multiple information sources: an augmented Gaussian process based algorithm

Contact Info

Product

Resources

About