Active Learning for Simulator Calibration

The Kennedy and O'Hagan (KOH) calibration framework uses coupled Gaussian processes (GPs) to meta-model an expensive simulator (first GP), tune its 'knobs' (calibration inputs) to best match observations from a real physical/field experiment and correct for any modeling bias (second GP) when predicting under novel field conditions (design inputs). There are well-established methods for placement of design inputs for data-efficient planning of a simulation campaign in isolation, i.e., without field data: space-filling, or via criteria like minimum integrated mean-squared prediction error (IMSPE). Analogues within the coupled GP KOH framework are mostly absent from the literature. Here we derive a novel, closed from IMSPE criteria for sequentially acquiring new simulator data in an active learning setting for KOH. We illustrate how acquisitions space-fill in design space, but concentrate in calibration space. Closed form IMSPE precipitates a closed-form gradient for efficient numerical optimization. We demonstrate that such acquisitions lead to a more efficient simulation campaign on benchmark problems, and conclude with a showcase on a motivating problem involving prediction of equilibrium concentrations of rare earth elements for a liquid-liquid extraction reaction.