Sequential Design and Rational Metamodeling

Hendrickx, W.; Dhaene, T.

doi:10.1109/wsc.2005.1574263

Cited by 22 publications

(14 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some model-based sampling strategies are listed in the third column. Model error sampling for instance, samples in areas with most disagreements between the model and the actual outputs [45]. Model-based methods typically pursue a specific goal: EDSD (Explicit Design Space Decomposition) [46] for instance is a sequential design method for refining the class boundary of an SVM model, whereas Probability of Feasibility [47] searches for areas which exceed a certain threshold making them suitable for sampling constrained areas.…”

Section: Existing Sequential Sampling Methodsmentioning

confidence: 99%

“…OUTPUT-BASED MODEL-BASED Low discrepancy sequences [48,49,50] Probability of Feasibility (PoF) [47] Sequentially Nested Latin Hypercubes [51,52,17] (F)LOLA-Voronoi (Regression) [16,53] Model error sampling (regression) [45] Montecarlo/Optimization Based [14,15] NeighbourhoodVoronoi (Classification) [54] D-and G-optimal designs [55,56] Voronoi-based [16] Explicit Design Space Decomposition (EDSD) [46] Random Sequential Exploratory Experimental Design (SEED) [42] (an output-based approach) and Probability of Feasibility [47] (a model-based approach). These methods are also applied to the test cases in Section 5.…”

Section: Input-basedmentioning

confidence: 99%

See 1 more Smart Citation

Adaptive classification under computational budget constraints using sequential data gathering

Herten

Couckuyt

Deschrijver

et al. 2016

Advances in Engineering Software

View full text Add to dashboard Cite

Classification algorithms often handle large amounts of labeled data. When a label is the result of a very expensive computer experiment (in terms of computational time), sequential selection of samples can be used to limit the overall cost of acquiring the labeled data. This paper outlines the concept of sequential design for classification, and the extension of an existing stateof-the-art research platform for surrogate modeling to handle classification problems with sequential design. The capabilities of the platform are illustrated on a number of use cases including real-world applications such as an ElectroMagnetic Compatibility (EMC) and a Computational Fluid Dynamics (CFD) problem. The CFD problem also illustrates how classification can be used together with regression techniques to solve multi-objective constrained optimization problems of complex systems.

show abstract

Section: Existing Sequential Sampling Methodsmentioning

confidence: 99%

Section: Input-basedmentioning

confidence: 99%

Adaptive classification under computational budget constraints using sequential data gathering

Herten

Couckuyt

Deschrijver

et al. 2016

Advances in Engineering Software

View full text Add to dashboard Cite

show abstract

“…Metamodelling techniques are increasingly used in simulation applications where full execution of a complex model for many different input conditions would be computationally expensive (Hendrickx and Dhaene, 2005). In effect, a metamodel aims to replace complex modelling of a deterministic system by a cheap numerical interpolating scheme which has been calibrated by sampling the original model's input and output (Piñeros Garcet et al, 2006).…”

Section: Methodsmentioning

confidence: 99%

A metamodelling implementation of a two-way coupled mesoscale-microscale flow model for urban area simulations

Tsegas

Barmpas

Douros

et al. 2011

IJEP

View full text Add to dashboard Cite

Systems of coupled prognostic mesoscale and microscale models have been used as a tool to accurately simulate flows around artificial structures and over densely-built urban areas. Typical implementations of such systems are based on a one-way coupling scheme, where the mesoscale model provides initial and boundary conditions for each off-line application of the microscale model. While very successful in predicting steady-state flows within specific local-scale areas, such schemes fail to account for feedbacks on the mesoscale flow induced by the presence of structures in smaller scales. Unfortunately, the large gap of spatial and temporal scales practically prohibits parallel on-line execution of the mesoscale and microscale models for any significant time interval. It is therefore necessary that a simplifying approach is adopted, where the microscale feedback is spatially and temporally upscaled to interact with parts of the mesoscale domain covering the urban area. In the present work a two-way coupled model system is developed, consisting of the prognostic mesoscale model MEMO and the microscale model MIMO. The microscale feedback on the mesoscale domain is simulated using a metamodelling approach, where the effect of local flows on the vertical profiles is estimated for representative urban areas of sizes up to a few hundred meters and used as calibration input for a set of interpolating metamodels. The feedback from the microscale metamodels is then introduced back in the mesoscale grid by means of Newtonian relaxation. As an illustrative application, simulations for the city of Athens, Greece during a multi-day period are presented. Effects of the microscale feedback on the mesoscale flow become evident both as a reduction of lower-level wind speeds in urban cells as well as an overall increase in turbulent kinetic energy production over densely-built areas.Key words: metamodelling, urbanisation, mesoscale, microscale, two-way coupling. INTRODUCTIONEfforts to introduce urban effects on mesoscale models have in general followed two different approaches. In the first one, corrections are applied in the mesoscale parameterisations within the lower computational layers, in an attempt to account for the specific characteristics of the urban canopy (Baklanov, 2004). Improved parameterisations include the effects of shadowing and radiation trapping as well as the urban heat island effect (Grimmond and Oke, 1999). The effect of urban canopy on the momentum flow and turbulent kinetic energy (TKE) production is usually parameterised as additional aerodynamic drag and production terms, respectively, introduced in the dynamical equations (Dupont et al., 2004). The so-called "urbanisation" approaches have the advantage of relatively low computational requirements and offer a large degree of flexibility in adjusting the relative strength of the forcing terms, allowing a fine-tuning to the particular characteristics of different urban morphologies. On the other hand, the number of introduced parameters is ofte...

show abstract

“…When the goal is an accurate model, space-filling sequential methods [18] can be used, or methods that analyse the responses [20], [36] and intermediate models [37] to increase sampling density in areas that are more difficult to model (poles, discontinuities, strong non-linearities etc). The benefit of the latter is illustrated in Figure 2: instead of purely spacefilling, the generated experimental design increases focus on a non-linear region.…”

Section: B Applications For Iotmentioning

confidence: 99%

Adaptive modeling and sampling methodologies for Internet of Things applications

Herten

Couckuyt

Deschrijver

et al. 2016

2016 18th Mediterranean Electrotechnical Conference (MELECON)

View full text Add to dashboard Cite

Abstract-The past few years several cloud services offer automated machine learning software. This enables non-experts to build sophisticated predictive models so they can focus on their area of expertise instead, and use these state-of-the-art machine learning techniques. These were the same principles that guided the development of the surrogate modeling (SUMO) toolbox to assist engineers during (virtual) product design and rapid prototyping with state-of-the-art machine learning methods. A proof of concept was developed, which exposes the technologies of the SUMO toolbox as a network service, offering them to the devices attached to the same network. Both the implementation of the service as well as the possibilities for Internet of Things are discussed.

show abstract

Sequential Design and Rational Metamodeling

Cited by 22 publications

References 14 publications

Adaptive classification under computational budget constraints using sequential data gathering

Adaptive classification under computational budget constraints using sequential data gathering

A metamodelling implementation of a two-way coupled mesoscale-microscale flow model for urban area simulations

Adaptive modeling and sampling methodologies for Internet of Things applications

Contact Info

Product

Resources

About