Defining "discrepancy model" when using the module of Bayesian inference in UqLab

Dear community,

I’m trying to use Bayesian inference to update my metamodel with the evidences from experimental observations. I understand the key step is to infer the unknow parameters based on the experimental data, so we can improve the prediction accuracy. My question is about the definition of model discrepancy. I wonder if Uqlab can realize similar bayesian calibration as mentioned by this paper: Kennedy MC, O’Hagan A. Bayesian calibration of computer models. J. Roy. Stat.
Soc. 2001;63:425–64.

I’m freshman in Bayesian calibration, so I’m not very familar with its procedure and implementation. And it seems the Uqlab manual on Bayesian calibration mainly emphasize the inference of the unknown parameters, while the description on the discrepancy model is few.

Hi @GPLai

Model bias terms are currently not supported out-of-the-box in the Bayesian inversion module. See also the discussion here.

However, adding this should be straightforward with the user-defined likelihood function feature. I suggest you have a look at the example uq_Example_Inversion_06_UserDefinedLikelihood and start from there.

Thanks. I’ll read the manual again.