{"id":"arxiv/2603.28638","url":"https://arxiv.org/abs/2603.28638","tags":["math.NA"],"score":0,"title":"Divergence-free Linearized Neural Networks: Integral Representation and Optimal Approximation Rates","domain":"math.NA","source":"arxiv","authors":["Juncai He","Xinliang Liu","Zitong Tian"],"reviews":0,"subject":"math.NA","summary":"This paper studies the numerical approximation of divergence-free vector fields by linearized shallow neural networks, also referred to as random feature models or finite neuron spaces. Combining the stable potential lifting for divergence-free fields with the scalar Sobolev integral representation theory via ReLU$^k$ networks, we derive a core integral representation of divergence-free Sobolev vector fields through antisymmetric potentials parameterized by linearized ReLU$^k$ neural networks. This representation, together with a quasi-uniform distribution argument for the inner parameters,…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-03-30","reactions":{}}