Improving trust in numerical simulations
Argonne Mathematics and Computer Science researchers have authored a white paper on improving trust in numerical simulations.
Trust is a serious and insufficiently recognized problem.
Definitions differ, both systematic and unsystematic; unintentional and malicious corruptions occur; and detection can be difficult and costly.
The white paper investigates key aspects of trust that users can give to the results of numerical simulations.
This white paper investigates several key aspects of the trust that a user can give to the results of numerical simulations and scientific data analytics. In this document, the notion of trust is related to the integrity of numerical simulations and data analytics applications. This white paper complements the DOE ASCR report on Cyber Security for Scientific Computing Integrity by:
- exploring the sources of trust loss;
- reviewing the definitions of trust in several areas;
- providing numerous cases of result alteration, some of them leading to catastrophic failures;
- examining the current notion of trust in numerical simulation and scientific data analytics;
- providing a gap analysis; and
- suggesting two important research directions and their respective research topics.
The white paper is an open document. The authors welcome additions, including examples of corruptions and techniques for improving trust.
Franck Cappello, Emil Constantinescu, Paul Hovland, Tom Peterka, Carolyn Phillips, Marc Snir and Stefan Wild, “Improving the Trust in Results of Numerical Simulations and Scientific Data Analytics,” MCS Report, ANL/MCS-TM-352.pdf, Published Online April 2015.