Discovery and exclusion limits without binning using machine-learned likelihoods

Febrero 9, 2023
De 3:00pm hasta 4:00pm

IFT Seminar Room/Red Room

Specialist level
Speaker: 
Andres D. Perez
Institution: 
Instituto de Física Teórica UAM-CSIC
Location&Place: 

IFT Seminar Room/Red Room

Abstract: 
Machine-Learned Likelihood is a method that, by combining the power of current machine-learning techniques with the likelihood-based inference tests used in traditional analyses, allows to estimate the experimental sensitivity of high-dimensional data sets. Based on supervised learning techniques, the addition of Kernel Density Estimators avoids the need to bin the classifier output in order to extract the resulting one-dimensional signal and background probability density functions. To explore the potential of the method, first we apply it to toy models of multivariate Gaussian distributions, where the true probability distribution functions are known. Then we test it in the search for new physics at the LHC considering two cases: a W' in dijet final states and a Z' decaying into lepton pairs.