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

Mazumdar, Atanu; Chugh, Tinkle; Miettinen, Kaisa; López-Ibáñez, Manuel

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

Cited by 13 publications

(12 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The design task can be more expensive numerically in a system-level design of a rotating machine, where the electrical machine and the control are designed together. To reduce the computational demand of these tasks, a wide variety of meta-modeling techniques can be employed to evaluate the fitness function and avoid complex numerical simulations [48,196,197]: response surface methodologies [198,199], multilayer perceptron [194,[200][201][202][203], radial basis function neural networks [204,205], kriging models [206,207], Gaussian processes [208], support vector machines [209][210][211], fuzzy logic [212].…”

Section: Methods For Computational Cost Reductionmentioning

confidence: 99%

Robust Design Optimization and Emerging Technologies for Electrical Machines: Challenges and Open Problems

et al. 2020

View full text Add to dashboard Cite

The bio-inspired algorithms are novel, modern, and efficient tools for the design of electrical machines. However, from the mathematical point of view, these problems belong to the most general branch of non-linear optimization problems, where these tools cannot guarantee that a global minimum is found. The numerical cost and the accuracy of these algorithms depend on the initialization of their internal parameters, which may themselves be the subject of parameter tuning according to the application. In practice, these optimization problems are even more challenging, because engineers are looking for robust designs, which are not sensitive to the tolerances and the manufacturing uncertainties. These criteria further increase these computationally expensive problems due to the additional evaluations of the goal function. The goal of this paper is to give an overview of the widely used optimization techniques in electrical machinery and to summarize the challenges and open problems in the applications of the robust design optimization and the prospects in the case of the newly emerging technologies.

show abstract

Section: Methods For Computational Cost Reductionmentioning

confidence: 99%

Robust Design Optimization and Emerging Technologies for Electrical Machines: Challenges and Open Problems

et al. 2020

View full text Add to dashboard Cite

show abstract

“…Generally speaking, as described in Section II-A, DDEAs can be classified into two categories: 1) offline DDEAs and 2) online DDEAs. As a number of DDEAs are proposed for solving multiobjective [32], [33] or manyobjective problems [34], the following contents will also clarify their multi-/many-objective characteristics when surveying them among the offline and online DDEAs.…”

Section: B Related Workmentioning

confidence: 99%

Boosting Data-Driven Evolutionary Algorithm With Localized Data Generation

Zhan

Wang

et al. 2020

IEEE Trans. Evol. Computat.

119

View full text Add to dashboard Cite

By efficiently building and exploiting surrogates, data-driven evolutionary algorithms (DDEAs) can be very helpful in solving expensive and computationally intensive problems. However, they still often suffer from two difficulties. First, many existing methods for building a single ad hoc surrogate are suitable for some special problems but may not work well on some other problems. Second, the optimization accuracy of DDEAs deteriorates if available data are not enough for building accurate surrogates, which is common in expensive optimization problems. To this end, this article proposes a novel DDEA with two efficient components. First, a boosting strategy (BS) is proposed for self-aware model managements, which can iteratively build and combine surrogates to obtain suitable surrogate models for different problems. Second, a localized data generation (LDG) method is proposed to generate synthetic data to alleviate data shortage and increase data quantity, which is achieved by approximating fitness through data positions. By integrating the BS and the LDG, the BDDEA-LDG algorithm is able to improve model accuracy and data quantity at the same time automatically according to the problems at hand. Besides, a tradeoff is empirically considered to strike a better balance between the effectiveness of surrogates and the time cost for building them. The experimental results show that the proposed BDDEA-LDG algorithm can generally outperform both traditional methods without surrogates and other state-of-the-art DDEA son widely used benchmarks and an arterial traffic signal timing real-world optimization problem. Furthermore, the proposed BDDEA-LDG algorithm can use only about 2% computational budgets of traditional methods for producing competitive results.

show abstract

“…Then, the empirical model (on which evaluating the fitness function is computationally inexpensive) is used to predict promising new solutions [157]. Current functional approximation models include Polynomials (response surface methodologies [47,172]), artificial neural networks (e.g., multi-layer perceptrons (MLPs) [5], radial-basis function (RBF) networks [2,177], Gaussian processes [15,179], support vector machines [3,4,176] and Kriging [111,126] models. Although frequently used in engineering applications, surrogate methods can normally be adopted only in problems of low dimensionality, which is an important limitation when dealing with real-world MOPs.…”

Section: Dealing With Expensive Objective Functionsmentioning

confidence: 99%

Evolutionary multiobjective optimization: open research areas and some challenges lying ahead

Coello¹,

Brambila²,

Gamboa³

et al. 2019

Complex Intell. Syst.

121

View full text Add to dashboard Cite

Evolutionary multiobjective optimization has been a research area since the mid-1980s, and has experienced a very significant activity in the last 20 years. However, and in spite of the maturity of this field, there are still several important challenges lying ahead. This paper provides a short description of some of them, with a particular focus on open research areas, rather than on specific research topics or problems. The main aim of this paper is to motivate researchers and students to develop research in these areas, as this will contribute to maintaining this discipline active during the next few years.

show abstract

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

Cited by 13 publications

References 24 publications

Robust Design Optimization and Emerging Technologies for Electrical Machines: Challenges and Open Problems

Robust Design Optimization and Emerging Technologies for Electrical Machines: Challenges and Open Problems

Boosting Data-Driven Evolutionary Algorithm With Localized Data Generation

Evolutionary multiobjective optimization: open research areas and some challenges lying ahead

Contact Info

Product

Resources

About