Case studies in Gaussian process modelling of computer codes

Kennedy, Marc C.; Anderson, C. W.; Conti, Stefano; O’Hagan, Anthony

doi:10.1016/j.ress.2005.11.028

Cited by 113 publications

(74 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the output from SWAN is generally not a linear function of the inputs, these traditional look-up tables can be inefficient and require a large number of design point simulations. There has, however, been extensive research into more sophisticated interpolation techniques, in particular Gaussian process emulators (GPEs) (Kennedy et al, 2006), for example. These more sophisticated approaches have been shown to be efficient when used in the context of wave transformation modelling (Camus et al, 2011a).…”

Section: Swan Emulationmentioning

confidence: 99%

Multivariate extreme value modelling of sea conditions around the coast of England

Gouldby

Wyncoll

Panzeri

et al. 2017

Proceedings of the Institution of Civil Engineers - Maritime En

View full text Add to dashboard Cite

It is widely recognised that coastal flood events can arise from combinations of extreme waves and sea levels. For flood risk analysis and the design of coastal structures it is therefore necessary to assess the joint probability of the occurrence of these variables. Traditional methods have involved the application of joint probability contours, defined in terms of extremes of sea conditions that can, if applied without correction factors, lead to the underestimation of flood risk and under-design of coastal structures. This paper describes the application of a robust multivariate statistical model to analyse extreme offshore waves, wind and sea levels around the coast of England. The approach described here is risk based in that it seeks to define extremes of response variables directly, rather than the joint extremes of sea conditions. The output of the statistical model comprises a Monte Carlo simulation of extreme events. These distributions of extreme events have been transformed from offshore to nearshore using a statistical emulator of a wave transformation model. The resulting nearshore extreme sea condition distributions have the potential to be applied for a range of purposes. The application is demonstrated using two structures located on the south coast of England. Notation

show abstract

Section: Swan Emulationmentioning

confidence: 99%

Multivariate extreme value modelling of sea conditions around the coast of England

Gouldby

Wyncoll

Panzeri

et al. 2017

Proceedings of the Institution of Civil Engineers - Maritime En

View full text Add to dashboard Cite

show abstract

“…Emulators are usually constructed using regression, with coefficients calibrated from an ensemble of simulations with different parameter settings (e.g. Murphy et al, 2004Murphy et al, , 2007Kennedy et al, 2006;Rougier and Sexton, 2007;Rougier et al, 2009;, though neural networks have also been used (e.g. Knutti et al, 2003;Piani et al, 2005;Sanderson et al, 2008).…”

Section: Emulationmentioning

confidence: 99%

Risk and uncertainty in hydrometeorological hazards

Edwards¹,

Challenor²

2013

Risk and Uncertainty Assessment for Natural Hazards

View full text Add to dashboard Cite

“…Each uncertain input parameter is represented as a probability distribution, and an emulator is fitted using multiple runs of the true model. In this case, the emulator is based on a powerful regression paradigm from machine learning called a Gaussian process [9]. From this emulator, statistical quantities relating to sensitivity and uncertainty can be inferred directly -for example, output uncertainty distributions and parameter sensitivities.…”

Section: Bayesian Sensitivity Analysismentioning

confidence: 99%

Aspects of Uncertainty Analysis for Large Nonlinear Computational Models

Worden

Becker

Battipede

et al. 2010

AMM

View full text Add to dashboard Cite

Abstract. This paper concerns the analysis of how uncertainty propagates through large computational models like finite element models. If a model is expensive to run, a Monte Carlo approach based on sampling over the possible model inputs will not be feasible, because the large number of model runs will be prohibitively expensive. Fortunately, an alternative to Monte Carlo is available in the form of the established Bayesian algorithm discussed here; this algorithm can provide information about uncertainty with many less model runs than Monte Carlo requires. The algorithm also provides information regarding sensitivity to the inputs i.e. the extent to which input uncertainties are responsible for output uncertainty. After describing the basic principles of the Bayesian approach, it is illustrated via two case studies: the first concerns a finite element model of a human heart valve and the second, an airship model incorporating fluid structure interaction.

show abstract

Case studies in Gaussian process modelling of computer codes

Cited by 113 publications

References 9 publications

Multivariate extreme value modelling of sea conditions around the coast of England

Multivariate extreme value modelling of sea conditions around the coast of England

Risk and uncertainty in hydrometeorological hazards

Aspects of Uncertainty Analysis for Large Nonlinear Computational Models

Contact Info

Product

Resources

About