Speaker
Description
The precise determination of the parton distribution functions (PDFs) of the proton is an essential ingredient for LHC analyses, including the upcoming High-Luminosity upgrade. PDFs are determined from a global scattering of hard scattering data taking as input unfolded low-dimensional binned cross- sections. In this work we demonstrate the feasibility of deploying neural simulation-based inference (NSBI) to constrain the proton PDFs from unbinned, high-dimensional, detector level observables taking into account all relevant experimental and theoretical systematic uncertainties. We develop procedures that allows for a detailed account of systematic uncertainties in the unbinned modeling of the high-dimensional data set and how that interplays with the PDF extraction. Adopting as proof-of-concept the determination of the large-x gluon PDFs from top quark pair production, preliminary results with our pipeline demonstrate significant sensitivity improvements as compared for traditional low-dimensional binned analyses.