Speaker
Description
Perfect lattice actions are renormalization-group improved actions that share their low energy properties with the continuum theory. This allows us to recover continuum physics even from very coarse lattices, with lattice artifacts being absent or at least notably suppressed. The difficulty in employing such actions in simulations is finding a suitable parameterization: in principle, they are allowed to have an arbitrarily complicated operator structure with arbitrarily many couplings.
In recent years, gauge-equivariant neural networks were successfully applied to this parameterization problem for the case of classically perfect actions, so called fixed-point actions. In this talk I will discuss progress on simulating these machine-learned fixed-point actions and illustrate how to go beyond classical, that is, toward quantum perfection.