Gauge-equivariant flow models for sampling in lattice field theories with pseudofermions

INSPIRE: 2116899
arXiv: 2207.08945
DOI: 10.1103/PhysRevD.106.074506

Authors: Abbott, Ryan, Albergo, Michael S., Boyda, Denis, Cranmer, Kyle, Hackett, Daniel C., Kanwar, Gurtej, Racanière, Sébastien, Rezende, Danilo J., Romero-López, Fernando, Shanahan, Phiala E., Tian, Betsy, Urban, Julian M.

Submitted: 18 July 2022

Subjects:

Journal reference: Phys.Rev.D 106 (7) 074506 (2022)

Citations: 68

Abstract

This work presents gauge-equivariant architectures for flow-based sampling in fermionic lattice field theories using pseudofermions as stochastic estimators for the fermionic determinant. This is the default approach in state-of-the-art lattice field theory calculations, making this development critical to the practical application of flow models to theories such as QCD. Methods by which flow-based sampling approaches can be improved via standard techniques such as even/odd preconditioning and the Hasenbusch factorization are also outlined. Numerical demonstrations in two-dimensional U(1) and SU(3) gauge theories with <math display="inline"><msub><mi>N</mi><mi>f</mi></msub><mo>=</mo><mn>2</mn></math> flavors of fermions are provided.