{"id":"arxiv/2610.38064","url":"https://arxiv.org/abs/2610.38064","tags":["math.NA","math.AP"],"score":0,"title":"Neural Stream Function Method for Electrical Impedance Tomography with Lipschitz Conductivities","domain":"math.NA","source":"arxiv","authors":["Xiaosheng Li"],"reviews":0,"subject":"math.NA","summary":"Electrical impedance tomography (EIT) reconstructs the conductivity inside an object from voltage and current measurements on its boundary. The resulting inverse problem is nonlinear and severely ill-posed. Neural networks offer a flexible computational approach, but training physics informed neural networks for EIT and relating their training loss to reconstruction accuracy remain challenging. We develop a novel constrained neural stream function method for planar EIT with Lipschitz conductivities. The method builds the conductivity constraints and prescribed boundary voltages into the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-27","reactions":{}}