{"id":"arxiv/2603.10078","url":"https://arxiv.org/abs/2603.10078","tags":["cs.LG","math.PR"],"score":0,"title":"Stochastic Port-Hamiltonian Neural Networks: Universal Approximation with Passivity Guarantees","domain":"cs.LG","source":"arxiv","authors":["Luca Di Persio","Matthias Ehrhardt","Youness Outaleb"],"reviews":0,"subject":"cs.LG","summary":"Stochastic port-Hamiltonian systems represent open dynamical systems with dissipation, inputs, and stochastic forcing in an energy based form. We introduce stochastic port-Hamiltonian neural networks, SPH-NNs, which parameterize the Hamiltonian with a feedforward network and enforce skew symmetry of the interconnection matrix and positive semidefiniteness of the dissipation matrix. For Itô dynamics we establish a weak passivity inequality in expectation under an explicit generator condition, stated for a stopped process on a compact set. We also prove a universal approximation result showing…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"discussed":0,"published":"2026-03-10","reactions":{}}