Speaker
Diaa Eddin Habibi
(Swansea University)
Description
In this talk, I will present our work on gauge equivariant diffusion models for simulating non-Abelian lattice gauge theories. Using the Metropolis-adjusted annealed Langevin algorithm (MAALA) together with lattice gauge equivariant convolutional neural networks (L-CNNs), we accurately reproduce the physics of two-dimensional U(2) and SU(2) gauge theories. Trained on a single Monte Carlo ensemble, our models generalise well to larger lattice sizes and stronger couplings with minimal accuracy loss. I will present results for Wilson loops and the topological susceptibility, demonstrating that diffusion models are a promising approach for lattice gauge theory beyond the Abelian case.
Authors
Dr
Andreas Ipp
(TU Wien)
David Müller
(TU Wien)
Diaa Eddin Habibi
(Swansea University)
Gert Aarts
(Swansea University)
Lingxiao Wang
(RIKEN)
Qianteng Zhu
(Shanghai Jiao Tong University)
Thomas Ranner
(TU Wien)
Prof.
Wei Wang
(Shanghai Jiao Tong University)