Lattice gauge symmetry in neural networks
This work provides a domain-specific solution for researchers in lattice gauge theory, offering improved performance in symmetry-preserving neural networks.
The paper introduces lattice gauge equivariant convolutional neural networks (L-CNNs) to address machine learning problems in lattice gauge theory by exactly preserving gauge symmetry, showing that L-CNNs achieve higher accuracy and better generalizability compared to non-equivariant CNNs in non-linear regression tasks.
We review a novel neural network architecture called lattice gauge equivariant convolutional neural networks (L-CNNs), which can be applied to generic machine learning problems in lattice gauge theory while exactly preserving gauge symmetry. We discuss the concept of gauge equivariance which we use to explicitly construct a gauge equivariant convolutional layer and a bilinear layer. The performance of L-CNNs and non-equivariant CNNs is compared using seemingly simple non-linear regression tasks, where L-CNNs demonstrate generalizability and achieve a high degree of accuracy in their predictions compared to their non-equivariant counterparts.