Extraction of multiparton interactions from ALICE pp collisions data using machine learning

Over the last years, Machine Learning (ML) methods have been successfully applied to a wealth of problems in high-energy physics. In this work, we discuss the extraction of the average number of Multiparton Interactions (⟨Nmpi⟩) from minimum-bias pp data at LHC energies using a regression based on Boosted Decision Trees (BDT). Using the available ALICE data on transverse momentum spectra as a function of multiplicity, we report that for minimum-bias pp collisions at √s = 7 TeV the average Nmpi is 3.98 ± 1.01, which complements our previous results for pp collisions at √s = 5.02 and 13 TeV. The comparisons indicated a modest center-of-mass energy dependence of ⟨Nmpi⟩. The study is further extended extracting the multiplicity dependence of Nmpi for the three center-of-mass energies. These results are qualitatively consistent with the existing ALICE measurements sensitives to Multiparton Interactions (MPI). Through the ML method applied to pp collisions at √s = 13 TeV, we also show that computing the multiplicity in the forward region the extraction of Nmpi is improved. This result opens the possibility to extract the number of MPI event-by-event, and in this way study the particle production as a function of MPI. Our results provide additional evidence of the presence of MPI in hadronic interactions and can help to the understanding of the heavy-ion-like behaviour observed in pp collisions data.