13th International Conference on the Exact Renormalization Group 2026 (ERG2026)

Constructing UV-critical hypersurfaces using machine learning

1 Sept 2026, 15:31

1m

Large Lecture Theatre (Jubilee Building)

Poster Poster presentations

Speaker

Stijn Hennissen (Radboud University)

Description

We provide a novel computational method for constructing the UV-critical surface along the lines of [arXiv:2403.08541]. Using physics informed neural networks to learn flow invariants, we recover the UV-critical surface all at once. The method is successfully applied to a UV-critical surface for an asymptotically safe scalar-tensor theory, where we predict the IR effective field theory. We argue that these types of networks are highly efficient computational tools extracting physics from flow equations.