Design of Experiments for Model Discrimination Hybridising Analytical and Data-Driven Approaches

  • 2018-02-12 16:34:06
  • Simon Olofsson, Marc Peter Deisenroth, Ruth Misener
  • 12


Healthcare companies must submit pharmaceutical drugs or medical devices toregulatory bodies before marketing new technology. Regulatory bodies frequentlyrequire transparent and interpretable computational modelling to justify a newhealthcare technology, but researchers may have several competing models for abiological system and too little data to discriminate between the models. Indesign of experiments for model discrimination, the goal is to design maximallyinformative physical experiments in order to discriminate between rivalpredictive models. Prior work has focused either on analytical approaches,which cannot manage all functions, or on data-driven approaches, which may havecomputational difficulties or lack interpretable marginal predictivedistributions. We develop a methodology introducing Gaussian process surrogatesin lieu of the original mechanistic models. We thereby extend existing designand model discrimination methods developed for analytical models to cases ofnon-analytical models in a computationally efficient manner.


Introduction (beta)



Conclusion (beta)