Neural Networks as Surrogate Models for Measurements in Optimization Algorithms

Holeňa, Martin; Linke, David; Rodemerck, Uwe; Bajer, Lukáš

doi:10.1007/978-3-642-13568-2_25

Cited by 10 publications

(5 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This computational time advantage would be even more impressive if the time to perform a single simulation was longer. Previous research has also clearly illustrated the advantage of using neural networks as surrogate models for the optimization of processes [21,22]. In this investigation, the number of evaluations using the phenomenological model for circumscribing the Pareto domain is given in Table 3.…”

Section: Resultsmentioning

confidence: 99%

Artificial Neural Networks as Metamodels for the Multiobjective Optimization of Biobutanol Production

2018

View full text Add to dashboard Cite

Process optimization using a physical process or its comprehensive model often requires a significant amount of time. To remedy this problem, metamodels, or surrogate models, can be used. In this investigation, a methodology for optimizing the biobutanol production process via the integrated acetone-butanol-ethanol (ABE) fermentation-membrane pervaporation process is proposed. In this investigation, artificial neural networks (ANNs) were used as metamodels in an attempt to reduce the time needed to circumscribe the Pareto domain and identify the best optimal operating conditions. Two different metamodels were derived from a small set of operating conditions obtained from a uniform experimental design. The first series of metamodels were derived to entirely replace the phenomenological model of the butanol fermentation process by representing the relationship that exists between five operating conditions and four performance criteria. The second series of metamodels were derived to estimate the initial concentrations under steady-state conditions for the eight chemical species within the fermenter in order to expedite convergence of the process simulator. The first series of metamodels led to an accurate Pareto domain and reduced the computation time to circumscribe the Pareto domain by a factor of 2500. The second series of metamodels led to only a small reduction of computation time (a factor of approximately 2) because of the inherently slow convergence of the overall fermentation process.

show abstract

Section: Resultsmentioning

confidence: 99%

Artificial Neural Networks as Metamodels for the Multiobjective Optimization of Biobutanol Production

2018

View full text Add to dashboard Cite

show abstract

“…Learning the model has been addressed using a variety of methods. Neural networks are popular for surrogate models (Holeňa et al, 2010) and have the advantage of allowing partial derivatives of the output to be calculated at each input variable, which is very useful for gradient based searches. The representation of multilayer perceptrons makes it difficult to explicitly fix or infer linkage structure, however, so that information is not available to guide linkage based search.…”

Section: Surrogate Model-based Optimisationmentioning

confidence: 99%

Learning and Searching Pseudo-Boolean Surrogate Functions from Small Samples

Swingler

2020

Evolutionary Computation

View full text Add to dashboard Cite

When searching for input configurations that optimise the output of a system, it can be useful to build a statistical model of the system being optimised. This is done in approaches such as surrogate model-based optimisation, estimation of distribution algorithms, and linkage learning algorithms. This article presents a method for modelling pseudo-Boolean fitness functions using Walsh bases and an algorithm designed to discover the non-zero coefficients while attempting to minimise the number of fitness function evaluations required. The resulting models reveal linkage structure that can be used to guide a search of the model efficiently. It presents experimental results solving benchmark problems in fewer fitness function evaluations than those reported in the literature for other search methods such as EDAs and linkage learners.

show abstract

“…Recently, popular applications of machine learning in atmospheric chemistry and physics include quantitative structure-activity relationship (QSAR) models that map molecular structures to compound properties as an alternative to time-consuming laboratory experiments or quantum mechanical calculations (Lu et al, 2021;Lumiaro et al, 2021;Galeazzo and Shiraiwa, 2022;Krüger et al, 2022;Xia et al, 2022). Holeňa et al (2010) used surrogate models in computationally costly evolutionary optimization and successfully enhanced this approach with the application of NNs. Tripathy and Bilionis (2018) used an NN to create surrogate models for expensive high-dimensional uncertainty quantification.…”

Section: Introductionmentioning

confidence: 99%

Accelerating models for multiphase chemical kinetics through machine learning with polynomial chaos expansion and neural networks

et al. 2023

View full text Add to dashboard Cite

Abstract. The heterogeneous chemistry of atmospheric aerosols involves multiphase chemical kinetics that can be described by kinetic multi-layer models (KMs) that explicitly resolve mass transport and chemical reactions. However, KMs are computationally too expensive to be used as sub-modules in large-scale atmospheric models, and the computational costs also limit their utility in inverse-modeling approaches commonly used to infer aerosol kinetic parameters from laboratory studies. In this study, we show how machine learning methods can generate inexpensive surrogate models for the kinetic multi-layer model of aerosol surface and bulk chemistry (KM-SUB) to predict reaction times in multiphase chemical systems. We apply and compare two common and openly available methods for the generation of surrogate models, polynomial chaos expansion (PCE) with UQLab and neural networks (NNs) through the Python package Keras. We show that the PCE method is well suited to determining global sensitivity indices of the KMs, and we demonstrate how inverse-modeling applications can be enabled or accelerated with NN-suggested sampling. These qualities make them suitable supporting tools for laboratory work in the interpretation of data and the design of future experiments. Overall, the KM surrogate models investigated in this study are fast, accurate, and robust, which suggests their applicability as sub-modules in large-scale atmospheric models.

show abstract

Neural Networks as Surrogate Models for Measurements in Optimization Algorithms

Cited by 10 publications

References 14 publications

Artificial Neural Networks as Metamodels for the Multiobjective Optimization of Biobutanol Production

Artificial Neural Networks as Metamodels for the Multiobjective Optimization of Biobutanol Production

Learning and Searching Pseudo-Boolean Surrogate Functions from Small Samples

Accelerating models for multiphase chemical kinetics through machine learning with polynomial chaos expansion and neural networks

Contact Info

Product

Resources

About