{"id":"arxiv/2602.08374","url":"https://arxiv.org/abs/2602.08374","tags":["stat.ML","cs.LG","math.PR","math.ST"],"score":0,"title":"Schrödinger bridge problem via empirical risk minimization","domain":"stat.ML","source":"arxiv","authors":["Denis Belomestny","Alexey Naumov","Nikita Puchkin","Denis Suchkov"],"reviews":0,"subject":"stat.ML","summary":"We study the Schrödinger bridge problem when the endpoint distributions are available only through samples. Classical computational approaches estimate Schrödinger potentials via Sinkhorn iterations on empirical measures and then construct a time-inhomogeneous drift by differentiating a kernel-smoothed dual solution. In contrast, we propose a learning-theoretic route: we rewrite the Schrödinger system in terms of a single positive transformed potential that satisfies a nonlinear fixed-point equation and estimate this potential by empirical risk minimization over a function class. We…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-02-09","reactions":{}}