Speaker
Description
Gauge fixing is an essential step in lattice QCD calculations, particularly for the study of gauge-dependent observables. Traditional iterative gauge-fixing algorithms are computationally expensive and can suffer from critical slowing down near fixed points, leading to increasing computational costs on large lattices. We present a machine-learning framework for lattice gauge fixing, in which gauge transformation matrices are constructed using Wilson lines through a convolutional neural network. As a test case, we focus on Coulomb gauge fixing, while the framework can be readily extended to other gauge-fixing conditions. We investigate the gauge-fixing performance and computational efficiency of different trained models and identify network architectures that retain competitive performance while reducing computational cost, particularly for large lattice volumes. Preliminary results for SU(3) gauge ensembles demonstrate that suitable network architectures can provide an efficient and scalable alternative to conventional iterative gauge-fixing methods.