BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:An exact framework for uncertainty quantification in Monte Carlo s
imulation
DTSTART;VALUE=DATE-TIME:20131015T152500Z
DTEND;VALUE=DATE-TIME:20131015T154500Z
DTSTAMP;VALUE=DATE-TIME:20201126T133657Z
UID:indico-contribution-1512683@indico.cern.ch
DESCRIPTION:Speakers: Saracco Paolo (INFN Genova (Italy))\nUncertainty Qua
ntification (UQ) addresses the issue of predicting non-statistical errors
affecting the results of Monte Carlo simulations\, deriving from uncertain
ties in the physics data and models they embed. In HEP it is relevant to p
article transport in detectors\, as well as to event generators.\n \nWe su
mmarize recent developments\, which have established the mathematical grou
nd of an exact framework for UQ calculation. This study assessed that in t
he case of a single uncertainty and under wide hypotheses a simple general
relation exists\, which relates the probability density function (PDF) of
the input to Monte Carlo simulation\, and of the output it produces. This
result has been empirically verified in a conceptually simplified Monte C
arlo simulation environment.\n \nIn this contribution we address the probl
em of extending this approach to the multi-variate case. A typical scenari
o in this context consists of predicting the dependence of simulation resu
lts on input cross section tabulations. We show that for a wide class of p
robability distributions of the input unknowns it is possible to determine
analytically the expected output PDF for any required observable\, both i
n the case of independent variations and in the case of linear correlation
s among the input variables. This class includes normal distributions\, fl
at (and in general finite interval distributions)\, and all the Levy stabl
e distributions - a four parameter family of heavy-tailed distributions\,
which includes the Breit-Wigner one.\n \nFor all these distributions it is
possible to evaluate exactly the confidence intervals for the physical ob
servables of experimental interest produced by the Monte Carlo simulation.
\n \nThis is a powerful environment to perform UQ in many physical cases o
f interest to HEP and low energy nuclear physics experiments.\n \nWe prese
nt the mathematical methods for uncertainty quantification and some applic
ations to relevant use cases.\n\nhttps://indico.cern.ch/event/214784/contr
ibutions/1512683/
LOCATION:Amsterdam\, Beurs van Berlage Berlagezaal
URL:https://indico.cern.ch/event/214784/contributions/1512683/
END:VEVENT
END:VCALENDAR