Machine learning approaches for measurements and calibrations in the context of a precision measurement of the W mass

by Dr Niccolò Foppiani

Tuesday 15 Jan 2019, 15:00 → 17:00 Europe/Rome

aula 131

Given the small theoretical uncertainty on the W mass prediction, and the huge amount of data collected by the LHC so far, a precise measurement of this quantity is of primary interest. Such a measurement may unveil new physics effects by showing that the standard model is inconsistent at a quantum (loop) level. Precision measurements of the W mass have been already performed by the the CDF and D0 collaborations and more recently by the ATLAS collaboration. However, in order to overcome the theoretical precision, of the order of 10^4, novel methods to understand and control the systematic uncertainties are needed.
In this talk I will show a novel approach that has been developed to measure and calibrate the hadronic recoil, one of the main ingredients of the measurement. This approach is based on machine learning techniques, that have been studied, understood and tested in a deep way, before carrying out the result.
Within this result, it has been demonstrated that the systematic uncertainty due to an imperfect knowledge of the W transverse momentum can be reduced by a factor three with respect to the standard method used in the past, without adding additional systematic bias.

2019-01-15 Seminario INFN.pdf