Dresden 2014 – wissenschaftliches Programm
Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
DY: Fachverband Dynamik und Statistische Physik
DY 28: Extreme Events
DY 28.5: Vortrag
Donnerstag, 3. April 2014, 10:30–10:45, ZEU 146
Improving Predictions using the Crooks Fluctuation Theorem — •Julia Gundermann1, Stefan Siegert2, and Holger Kantz1 — 1Max Planck Institut für Physik komplexer Systeme, Dresden, Germany — 2University of Exeter, United Kingdom
The Crooks fluctuation theorem is a relation from non-equilibrium thermodynamics, quantifying the amount of work produced in a process more exactly than the second law inequality by giving an exact equation for the work's distribution.
We take this equation as a constraint for the distribution of a random variable and ask the question: Given a finite data set drawn from such a distribution, how can we improve the estimate of this variable to exceed a certain threshold compared to the event frequency deduced from the data set? Using the knowledge about Crooks' relation we propose a forecast that will prove to be "better" than simple counting on the data set. We measure the notion of "better" in terms of the Brier score. Studies of parameters such as exceedance threshold and data set size are presented.