Monday, October 28, 2019 - 12:50pm to 1:50pm
As science and industry become more and more dependent on complex numerical models for making predictions and designing structures and devices, how can we know whether a model is an accurate representation of reality? Sez Atamturkur's work aims to measure uncertainty and systematic error in these models. Modelers often focus on building logical algorithms and incorporating massive amounts of data, but Atamturkur says that process can introduce systematic error or bias, which are exacerbated when models are coupled. She and colleagues have focused on rooting out the nature of error or bias and representing it in a physically meaningful way so that models can be validated. She will show how the work is being used to validate models used to test medical equipment.