Joint Research Conference

June 24-26, 2014

He output from a computer model of a system often depends on a set of parameters which are unknown to the designer of the model but can be estimated by collecting physical data. We investigate the use of physical data to distinguish aspects of a computer model that change under an ambiguous or wide reaching treatment performed on a system. By describing the treatment effect via an interpretable model we are able to communicate important contextual clues. Ideas are illustrated through a study of the action potential of ventricular myocytes where the computer model has a high dimensional parameter space.