{"id":"arxiv/2608.19331","url":"https://arxiv.org/abs/2608.19331","tags":["hep-th","cond-mat.dis-nn","cs.LG"],"score":0,"title":"Deep neural networks as lattice gauge theories","domain":"hep-th","source":"arxiv","authors":["Ro Jefferson","Shradha Ramakrishnan"],"reviews":0,"subject":"hep-th","summary":"We modify the NN/QFT duality [1] to incorporate the layerwise permutation symmetry of the network, resulting in a $(0\\!+\\!1)$-dimensional lattice gauge theory, in which each layer of $N$ neurons acts as an $N$-component lattice site, and the weight matrices play the role of gauge fields living on the links. In this framework, we compute the tree-level neuron-neuron propagator which describes the evolution of layer variance in the network, and develop the Feynman diagram machinery to compute interactions in the perturbative expansion in $1/N$. In particular, we obtain a recursive expression…","category":"research","comments":0,"disputed":false,"division":"physics","mentions":0,"discussed":0,"published":"2026-08-19","reactions":{}}