Berlin 2018 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
SYMS: Symposium Data-driven Methods in Molecular Simulations of Soft-Matter Systems
SYMS 1: Data-driven Methods in Molecular Simulations of Soft-Matter Systems
SYMS 1.1: Hauptvortrag
Montag, 12. März 2018, 15:00–15:30, H 0105
Stochastic numerical algorithms: from molecular dynamics to big data analytics — •Benedict Leimkuhler — University of Edinburgh, Edinburgh, UK
I will discuss the interplay between the methods of molecular/particle simulation and algorithms for statistical inference, as used for example for inference of model properties and parameter selection to describe a large data set. On the one hand, I will describe some of the increasingly sophisticated methods available for sampling molecular conformations, including constrained stochastic methods and other schemes which make use of variable temperature. I will show that many of these methods have analogous applications in data science, where they can help in the discovery of robust parameterisations or to address challenges due to problem structure. When the data set itself is generated by molecular dynamics trajectory simulation, the resulting procedures offer prospects for coarse-graining and enhanced sampling of highly complex systems.