Searching for Unexpected New Physics at the LHC with Machine Learning

New Physics Learning Machine (NPLM) is a machine-learning based strategy to detect data departures from a Reference hypothesis (the Standard Model), with no prior bias on the source of the discrepancy responsible for it. The main idea behind the method is to approximate the log- likelihood-ratio hypothesis test parametrising the data distribution with a universal approximating function, and solving its maximum-likelihood fit as a machine-learning problem, with a customised loss function. The method returns a p-value that measures the compatibility of the data with the Reference model. A strategy to account for the uncertainties of the Reference hypothesis has been developed, opening up the way to the application of NPLM to new physics searches at the LHC experiments. Beside that, the most interesting potential applications of NPLM are validation of new Monte Carlo event generators and data quality monitoring. Using efficient large-scale implementations of kernel methods as universal approximators, the NPLM algorithm can be deployed on a GPU-based data acquisition system and be exploited to explore online the readout of an experimental setup. This would allow to identify detector malfunctions or, possibly, unexpected anomalous patterns in the data. One crucial advantage of the NPLM algorithm over most of the standard goodness-of-fit tests routinely used in many experiments is its capability of inspecting multi-dimensional problems, thus being sensitive to correlations. It also identifies the most discrepant region of the phase-space and it reconstructs the multidimensional data distribution, allowing for further inspection and interpretation of the results. The purpose of this thesis is to develop, test and deploy the NPLM strategy. After presenting its conceptual foundations and main properties, the NPLM algorithm is applied to a real new physics search and a data quality monitoring problem. For the former, we analyse the dimuon final state of proton-proton collision events collected by the CMS experiment at a center-of-mass energy of 13 TeV. For the latter, we monitor the readout of drift tubes chambers collecting cosmic muons at the INFN Legnaro national laboratory.