Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 11: Statistics and Dynamics of/on Networks (joint session BP/DY/SOE)
DY 11.5: Vortrag
Dienstag, 12. März 2013, 10:45–11:00, H47
High performance simulation and visualization of epidemics on complex networks — •Peter A. Kolski1,2, Thomas Selhorst1, Martin Clauss3, and Jörn Hoffmann3 — 1Friedrich-Loeffler-Institut, Wusterhausen, Germany — 2University of Potsdam, Germany — 3University of Leipzig, Germany
Dynamical processes on complex networks are a growing field of interest. Performing simulations on large system of this kind demand a high computational power. To handle dynamics on networks the NetEvo C++ library can assign dynamical systems to edges and nodes. Furthermore it solves these ODEs via the ODEint library and can perform heuristic optimization. We introduce an extension to NetEvo using OpenCL on GPUs. With this approach we achieve an increase of computational performance up to a factor of 100, compared to an optimized C++ code on a modern CPU. Additionally we developed a framework to visualize intermediate results and to perform instantaneous visual analytics. The software will be applied in epidemiology, simulating disease spread on trade networks by solving the SIR model*s ODEs. The modification of parameter in real-time and the immediate access to simulation results leads to intuitive insights into the behavior of epidemics on large complex networks.