Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks
Abstract
Gauge-equivariant graph neural networks generate ordered products of gauge links through local message passing, but the geometry of these internal structures is not captured by the receptive field alone. We investigate the statistical ensemble of closed lattice walks generated by recursive node- and edge-feature operations, independently of network training. Analytical and numerical results reveal a separation between rapid growth of contour length and much slower spatial expansion: deeper compositions generate increasingly long paths through repeated traversal of relatively compact regions. Loop expansions of fermionic observables provide a physical interpretation of this hierarchy, showing how spatial locality remains compatible with high-order dependence on nearby gauge fields. These findings identify complementary roles of network depth in extending spatial reach and enriching the representations constructed within that reach, providing a geometrical perspective on learning lattice gauge observables.
Related
- From Vacuum Geometry to Solution-Phase Electronic Properties and Thermodynamics via Equivariant Neural Networks
- Learning Low Modes of the Wilson Dirac Operator with Gauge-Equivariant Networks
- Gram--Wishart--Stiefel formulation of the N=2, large--d gauge theory in 1D: towards emergent Grassmannian geometry
- Accelerating dynamic polarizability calculations of organic molecules using equivariant graph neural networks
- Learning the self-energy through equivariant neural networks
BibTeX
@misc{rayat2026emergent,
title = {{Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks}},
author = {Ali Rayat and Gia-Wei Chern},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.18742},
eprint = {2610.18742},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.18742},
}
RIS
TY - UNPB
TI - Emergent Loop Geometry in Gauge-Equivariant Graph Neural Networks
AU - Ali Rayat
AU - Gia-Wei Chern
PY - 2026
DA - 2026/10/13
PB - arXiv
DO - 10.48550/arxiv.2610.18742
UR - https://arxiv.org/abs/2610.18742
AB - Gauge-equivariant graph neural networks generate ordered products of gauge links through local message passing, but the geometry of these internal structures is not captured by the receptive field alone. We investigate the statistical ensemble of closed lattice walks generated by recursive node- and edge-feature operations, independently of network training. Analytical and numerical results reveal a separation between rapid growth of contour length and much slower spatial expansion: deeper compositions generate increasingly long paths through repeated traversal of relatively compact regions. Loop expansions of fermionic observables provide a physical interpretation of this hierarchy, showing how spatial locality remains compatible with high-order dependence on nearby gauge fields. These findings identify complementary roles of network depth in extending spatial reach and enriching the representations constructed within that reach, providing a geometrical perspective on learning lattice gauge observables.
ER -