Abstract:
New Physics Learning Machine (NPLM) is a machine-learning based strategy to detect data departures from a Reference model, with no prior bias on the nature of the phenomenon responsible for it. The main idea behind the method is to approximate the likelihood-ratio test statistic directly from the data a with flexible machine learning model.
Using an efficient large-scale implementation of kernel methods as universal approximators, the NPLM algorithm can be deployed on a GPU-based data processing system and be exploited for quasi-online applications, like the monitoring of an experimental apparatus readout.
Moreover, a neural-networks-based version of NPLM allows to deal with the uncertainties that affect the Reference model predictions in physics analyses of collider data. This extended formulation of the algorithm directly builds on the specific maximum-likelihood-ratio treatment of uncertainties as nuisance parameters, that is routinely employed in high-energy physics for hypothesis testing.
In this talk I will outline the conceptual foundations of the NPLM algorithm, the main features concerning the regularization of the ML task and the procedure to account for systematic uncertainties. I will show some examples of NPLM applications to model-independent new physics searches and more general goodness of fit problems. I will conclude with an outlook of the ongoing work in the direction of improving the NPLM algorithm performances.
arXiv:2305.14137
arXiv:2303.05413
Eur. Phys. J. C 82, 879 (2022)
Eur. Phys. J. C 82, 275 (2022)
Eur. Phys. J. C 81, 89 (2021)
Phys. Rev. D 99, 015014
ZOOM LINK: https://uci.zoom.us/j/97010614727