Verso Funzioni di Distribuzione Partoniche Polarizzate al next-to-next-to-leading order

I present the first determination of polarised parton distribution functions (PDFs) at next-to-next-to-leading order (NNLO) accuracy in perturbative QCD. The analysis relies on polarised inclusive (DIS) and semi-inclusive (SIDIS) deep inelastic scattering data. PDFs are parametrised in terms of a neural network, trained on the included data sets using stochastic gradient descent and cross-validation optimisation methods. Uncertainties of the PDFs are estimated by means of the Monte Carlo sampling method, which accounts for all sources of experimental uncertainties provided by the experiments. Observables are computed in the framework of perturbative QCD up to NNLO accuracy, exploiting the fast-kernel method implemented in the APFEL++ library. Flavour sensitivity of the SIDIS measurements released by the HERMES and COMPASS experiments allows for a determination of a set of seven independent PDF combinations. The impact of some theoretical constraints adopted in the analysis is discussed, as well as the comparison with other available PDF sets. ​