Professor Peter Coveney, UCL
Host: Alin Elena - this seminar is part of CCP5 Summer School (summer2021.ccp5.ac.uk)
Wednesday 21st July at 16:15 - 17:15
Zoom Link: https://ukri.zoom.us/j/95413937942
Reliability and reproducibility in computational science
The objectivity of science is its crowning feature. Its stock-in-trade are experimental facts, observations and theories which do not depend on who reports them but rather on the notion that the same findings would be obtained by anyone else performing similar procedures. This is what is meant by scientific reproducibility. In the modern era of science, computers have come to play a central role. Computer simulation is a way of extracting useful information from theories and the models built using them. Such models are typically impossible to analyse without computers. They produce results which may be designed for comparison against existing experimental measurements; but they are also capable of making predictions for which no experimental data are available, including ahead of important events. Owing to the sophistication of modern science, such calculations often require powerful computers if results of any kind are to be forthcoming. And for situations in which it is thought that the theories and models are sufficiently accurate, one would like to use computer-based simulation in order to make actionable predictions—predictions whose credibility is sufficiently great that we can use them to make important decisions.
In this talk, I'll take a closer look at classical molecular dynamics through this prism. It is a computer simulation technique that is in widespread use across many areas of science, from physics and chemistry to materials, biology and medicine. The method continues to attract criticism due its oft-reported lack of re-producibility which is in part due to a failure to submit it to reliable uncertainty quantification (UQ). I'll tell you about a detailed UQ study of such codes and move on to discuss other scientific domains where such analyses are fundamental to the making of actionable decisions.