{"id":"arxiv/2608.08040","url":"https://arxiv.org/abs/2608.08040","tags":["math.PR","cs.LG"],"score":0,"title":"A cylindrical neural approximation theorem for conditional laws of McKean-Vlasov equations with common noise","domain":"math.PR","source":"arxiv","authors":["Nacira Agram","Reda Hmioui","Jan Rems"],"reviews":0,"subject":"math.PR","summary":"We introduce conditional cylindrical neural networks for approximating functionals of conditional laws in McKean-Vlasov equations with common noise. Fourier moments of the initial law and truncated signatures of the time augmented common noise are mapped by a mixture density network to a Gaussian mixture approximation of the conditional law. A cylindrical neural network then evaluates the target functional through analytic integrals against this predicted measure. Rough path well posedness and stability provide a conditional law map that is continuous in the initial distribution and the…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-08","reactions":{}}