Equivariant flow-based sampling for lattice gauge theory
This work addresses computational bottlenecks in physics simulations for researchers studying lattice gauge theories, offering a significant improvement in sampling efficiency.
The authors tackled the problem of efficiently sampling lattice gauge theories by developing a gauge-invariant flow-based machine learning algorithm, achieving orders of magnitude greater efficiency in sampling topological quantities near critical points compared to traditional methods like Hybrid Monte Carlo and Heat Bath.
We define a class of machine-learned flow-based sampling algorithms for lattice gauge theories that are gauge-invariant by construction. We demonstrate the application of this framework to U(1) gauge theory in two spacetime dimensions, and find that near critical points in parameter space the approach is orders of magnitude more efficient at sampling topological quantities than more traditional sampling procedures such as Hybrid Monte Carlo and Heat Bath.