26 July 2026 to 1 August 2026
University of Maryland, College Park
US/Eastern timezone

Renormalization Group inspired inverse blocking with conditional normalizing flows

30 Jul 2026, 15:20
20m
Benjamin Banneker B (Adele H. Stamp Student Union)

Benjamin Banneker B

Adele H. Stamp Student Union

3972 Campus Dr, College Park, MD 20742
Contributed talk Algorithms and artificial intelligence Algorithms and artificial intelligence

Speaker

Letizia Parato (University of Colorado - Boulder)

Description

Critical slowing down remains a major challenge for lattice simulations near continuous phase transitions, where the correlation length diverges. One possible strategy to address this issue is to generate large-volume configurations from smaller lattices, where local update algorithms remain efficient. In this setting, the renormalization group provides a natural framework: the coarse configuration constrains the infrared degrees of freedom, while only the missing ultraviolet modes need to be reconstructed. Rather than constructing a perfect action for a fixed blocking transformation, we investigate the complementary idea of constructing what we term a "perfect blocking" transformation, chosen such that the RG flow approximately preserves the target action. This allows coarse configurations to be generated from a known probability distribution and subsequently upscaled to a larger volume with an exact accept/reject correction. We present the formulation of this framework and its implementation for the two-dimensional $\phi^4$ theory. The conditional normalizing-flow reconstruction of the ultraviolet degrees of freedom is briefly outlined.

Authors

Anna Hasenfratz Ethan Neil Letizia Parato (University of Colorado - Boulder)

Presentation materials

There are no materials yet.