Bayesian inference for the uncertainty distribution of computer model outputs

By Jeremy E. Oakley, Anthony OʼHagan. 2002

380 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Statistics and Probability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Statistics and Probability

Citations in the corpus

Probabilistic sensitivity analysis of complex models: a Bayesian approach (2004, methodology): “To estimate the posterior uncertainty about each Si, a simulation method presented in Oakley and O’Hagan (2002) is used.”
Adaptive Gaussian Process Approximation for Bayesian Inference with Expensive Likelihood Functions (2017, methodology): “Given a real-valued function g ( x ), the GP or the Kriging method constructs a surrogate model of g ( x ) in a nonparameteric Bayesian regression framework [23, 15, 17].”
Hybrid Dirichlet mixture models for functional data (2009, background): “In computer modelling (e.g. Oakley and O’Hagan (2002)) and machine learning (Neal, 1997; Rasmussen and Williams, 2006) Gaussian process realizations are often used as a basis to model random functions.”
Uncertainty in prior elicitations: a nonparametric approach (2007, neutral): “Finally, given m, v and b∗, we can sample a density function f (·) at a finite number of values of θ from the posterior p{f (·) |D,m, v, b∗}, as in Oakley & O’Hagan (2002) the joint posterior distribution of {f (θ1),. .”
Survey of multifidelity methods in uncertainty propagation, inference, and optimization (2018, methodology): “Often, Gaussian-process models are used to attempt to correct model inadequacy [114, 115, 100, 157, 18].”
Bayesian analysis of computer code outputs: A tutorial (2006, methodology): “Theory for the whole UA uncertainty distribution is given by Oakley and O’Hagan [4] and for percentiles of that distribution by Oakley [9].”
Bayesian emulation of complex multi-output and dynamic computer models (2010, methodology): “The single-output Bayesian methodology elaborated by O’Hagan (1992); Oakley and O’Hagan (2002) is extended in Section 2 to enable the simultaneous emulation of a vector of outputs.”
Sequential design of computer experiments for the estimation of a probability of failure (2010, methodology): “This line of research has its roots in the field of design and analysis of computer experiments (see, e.g., Sacks et al., 1989; Currin et al., 1991; Welch et al., 1992; Oakley and O’Hagan, 2002, 2004; Oakley, 2004; Bayarri et al., 2007).”
Statistical emulation of climate model projections based on precomputed GCM runs (2014, methodology): “UsingGaussian processes to emulate computer models is attractive in many circumstances because it does not require the prior assumption of any particular parametric form for the relationship between inputs and outputs and provides an internally consistent approach to estimating the uncertainties of ”
A Framework for Validation of Computer Models (2007, background): “…signiﬁcant body of work in the computer modeling area is that addressing the importance and uncertainty of input variables and/or the corresponding output distributions (propagation of error) for example, Saltelli et al. (2000), Oakley and O’Hagan (2002), Oakley (2004), Oakley and O’Hagan (2004).”
Emulation of a complex global aerosol model to quantify sensitivity to uncertain parameters (2011, methodology): “The Gaussian process has been used to carry out uncer- tainty analysis (Haylock and O’Hagan, 1996; O’Hagan and Haylock, 1997; Oakley and O’Hagan, 2002) including methods for estimating the percentiles of the output uncertainty distribution.”