Alessandro Nada, "Scalable flow-based samplers for lattice gauge theory"

Monday 2 Mar 2026, 15:00 → 17:00 Europe/Berlin

Nußallee 14-16/1.022 (HISKP) - Seminar Room III (HISKP)

Flow-based samplers represent a promising candidate to successfully mitigate critical slowing down in lattice field theory simulations. In this talk we first review recent developments of this class of generative machine-learning models and the challenges that they face when applied to four-dimensional lattice gauge theories. Then, we analyze the interesting scaling properties that flow-based samplers manifest when combined with out-of-equilibrium simulations based on Jarzynski equality in the so-called Stochastic Normalizing Flows (SNFs) framework. Lastly, a strategy to mitigate topological freezing in the case of the SU(3) pure gauge theory using SNFs is outlined.