Automated uncertainty quantification for numerical solutions of partial differential equations (AUQ-PDE)
Minimising errors in computational science
Uncertainties and measurement errors permeate all fields of computational science. In the biomedical disciplines, the uncertainties inherent in data acquisition and processing pose a fundamental challenge in our era of patient-specific modelling and simulation. The quantification of such uncertainties and their implications is vital for the predictive capabilities of computer simulations.
In spite of its importance, the role of uncertainty quantification remains underdeveloped in these data-driven scientific fields. This can be attributed to a critical and problematic gap between clinicians, biomedical engineers, numerical-method and algorithm designers, and scientific software developers. Such gaps are not restricted to the biomedical domain; indeed, these gaps present a generic challenge in the field of computational science.
Developing generic software
The AUQ-PDE project (Automated uncertainty quantification for numerical solutions of partial differential equations) aims to bridge this gap by developing and integrating generic software components featuring a high degree of automation for uncertainty quantification in computational models. The usability of the software will be adapted for application in various domains through the core involvement of application domain specialists.
The software developed will allow scientists and engineers to build customised models rapidly and quickly equip these models with customised uncertainty quantification methods. In the longer term, the more widespread availability of biomedical simulation studies with quantified uncertainties will have a positive impact on the use of such studies for clinical practice. For instance, this project will strengthen the possibility of using patient-specific simulations as a diagnostic or treatment planning tool for cardiac diseases.
The project participants are fully committed to ensuring open access to all relevant project results for the scientific community and general public. The software components, including both program code and documentation, will be regularly released under an open source licence such as the Lesser GNU Public License (LGPL). Furthermore, the software will be made available to the scientific community via standard, Open Access hosting channels such as Bitbucket or Github. Subsequently, all research articles produced in association with the project will either be published in Open Access journals or as preprints made publicly available through electronic archives and distribution servers such as arxiv.org.
The project has received NOK 4 million in funding over a three-year period (1 April 2015–31 March 2018).
The institutions participating in the project are the Center for Biomedical Computing at Simula Research Laboratory, Norway; the Department of Mathematical Sciences at Chalmers University of Technology, Sweden; and the Department of Mathematics and Statistics at the University of Helsinki, Finland.