Advances in TMD phenomenology with Neural Networks
par
In recent years, the extraction of Transverse Momentum Distributions (TMDs) has entered a new phase in which methodological advancements are becoming as important as new data. In this talk, I will focus on two recent key developments in TMD phenomenology. First, I will present a new neural-network–based approach for parametrizing TMDs developed within the framework of the MAP Collaboration. This approach significantly enhances the flexibility of the parametrization and reduces functional biases, allowing for a more faithful reconstruction of intrinsic transverse momentum distributions. Second, I will discuss very recent progress toward combining experimental measurements with lattice-QCD calculations to constrain the nonperturbative part of TMD evolution. This emerging synergy offers a promising path to achieving a more controlled description of the nonperturbative behaviour of the Collins-Soper kernel, where phenomenology alone typically suffers from limited sensitivity. These methodological advances are expected to play a crucial role in enabling precision TMD studies for future facilities such as the Electron-Ion Collider (EIC) and the high-luminosity upgrade of the Large Hadron Collider (LHC).