arXiv Science⌕ Search

arXiv · 2611.20990

Transferable neural-network preconditioners for Wilson--Clover propagator inversions

Abstract

We investigate two single--layer neural-network architectures for accelerating four-dimensional Wilson--Clover propagator inversions on lattice ensembles in practical use; both transfer directly without retraining. The first is a gauge-covariant parallel transport convolution (PTC) network that directly reduces the number of generalized conjugate residual (GCR) solver iterations. The second is a CNN-like network used together with a multigrid-inspired algorithm. The two approaches can be combined to further reduce GCR iterations. PTC transfers without retraining across lattice volumes, lattice spacings, and pion masses while maintaining comparable acceleration, whereas the CNN-like network transfers directly from small- to large-volume ensembles with strong acceleration. A cost comparison accounting for both basis construction and solves shows that the proposed scheme is more advantageous for large-volume ensembles than for small-volume ones.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Shiyi Zhong, Yuanyi Zhang, Jian-Hui Zhang, Dian-Jun Zhao. 2026-11-16. Transferable neural-network preconditioners for Wilson--Clover propagator inversions. https://arxiv.science/abs/2611.20990

Citations retain the original authors and DOI and include this site’s record URL. If this discovery service contributed to your research, you can also cite arXiv Science by Indiciae: BibTeX · RIS.

KEEP EXPLORING

Related papers

Generative diffusion models for lattice field theory

Generative diffusion models offer a new route to ensemble generation in lattice field theory. The backward, or denoising, process underlying these models is a stochastic process with a learned, time-dependent drift, which leads to a clear connection with stochastic quantisation. After reviewing this correspondence, I turn to theories with a sign problem. There the complex Boltzmann weight obstructs direct sampling, but complex Langevin dynamics can sometimes evade the sign problem by generating configurations on a complexified manifold, distributed according to a real and semi-positive distribution. This distribution is not known in closed form. I discuss how this distribution can be learned from complex Langevin data using score-based and, in particular, energy-based diffusion models. The latter give direct access to the unnormalised distribution itself, which can subsequently be used as the weight in a standard Markov chain Monte Carlo simulation. Results are shown for a simple quartic model with a complex action.

hep-lat↗

Global analysis of the $ρ$ resonance in lattice QCD

We present a lattice-QCD study of the $ρ$ resonance in which we address the quark-mass and lattice-spacing dependence in a novel way using a global analysis of four $N_f=2+1$ ensembles. We parametrize discretization effects in the lattice energy levels and not in the scattering amplitude, which is defined in the continuum. We describe the amplitude using a $K$-matrix parameterization, which contains the light- and strange-quark mass dependence. By varying the parameterizations of quark-mass and lattice-spacing dependence, we examine the sensitivity of the physical-limit $ρ$ pole position to these choices and obtain the result $\sqrt{s}_ρ= 744(11)(7) - i\,69(7)(3)~\text{MeV}$ from a model average.

hep-lat↗

Finite temperature Sp(4) lattice gauge theory with matter fields

We perform the first finite-temperature lattice study of the Sp(4) gauge theory coupled to two Wilson-Dirac fermions transforming in the fundamental representation. Its continuum limit plays an important role in many extensions of the Standard Model for particle physics with composite dynamics. For example, it provides a compelling realisation of the strongly interacting massive particle (SIMP) paradigm for dark matter. If realised in nature, it might lead to a first-order thermal phase transition in the early universe, and produce a detectable relic stochastic background of gravitational waves (GW). Assessing whether future experiments are sensitive to the resulting GW power spectrum requires determining the strength of the transition. We scan the parameter space of this lattice theory with a fine mesh of choices of lattice coupling, for two small values of the temporal extent of the lattice, and several choices of spatial volume and fermion mass. We measure the Polyakov loop and its susceptibility on the ensembles generated, implement reweighting, and perform a finite volume analysis, both to classify the nature of the phase transition, and to estimate the critical coupling. We validate our algorithms and analysis procedures on the better understood Sp(4) Yang-Mills theory, by confirming the presence of a first-order phase transition. Upon coupling the theory to matter fields, in the presence of moderately light fermions, the Polyakov loop shows a clear crossover behaviour and does not detect singularities associated with a phase transition. We also uncover the first evidence of a first-order thermal phase transition in the case of the largest fermion masses available. This finding motivates future, large-scale studies aimed at characterising the transition in the continuum limit, for this and closely related theories with Sp(2N) gauge group.

hep-lat↗