Bereiche | Tage | Auswahl | Suche | Downloads | Hilfe

HK: Physik der Hadronen und Kerne

HK 48: Instrumentation und Anwendungen VII

HK 48.5: Vortrag

Donnerstag, 20. März 2003, 18:00–18:15, F

Improving Signal to Noise Ratio using Evolutionary Algorithms^∗ — •Rüdiger Berlich — Lehrstuhl für Experimentalphysik 1, Ruhr-Universität Bochum, Universitätsstr. 150, 44780 Bochum

Evolutionary Algorithms (EA) provide a generic means of finding solutions to almost arbitrary optimization problems. The signal-to-noise ratio of signals in particle physics analysis can be regarded as a direct function of the cuts applied to numeric and discrete input variables. Evolutionary Algorithms can therefore be used to improve the signal-to-noise ratio by finding suitable cuts. The talk gives examples of the applicability of EA to this problem. A new implementation of EA is presented that allows to perform this optimization in parallel on systems ranging from single-processor machines to clusters and the GRID.

^∗ supported by the BMB+F