{"id":"arxiv/2608.13201","url":"https://arxiv.org/abs/2608.13201","tags":["stat.ML","cs.LG","math.OC","math.ST"],"score":0,"title":"Sinkhorn Linearization and the Spectral Proxy: Unifying the Statistical and Algorithmic Theory of Feature-Parameterized Inverse Optimal Transport via a Single Spectral Sandwich","domain":"stat.ML","source":"arxiv","authors":["Han Dong","Jiaming Li","Yongqiang Gong","Ruixi Li","Yin Liu"],"reviews":0,"subject":"stat.ML","summary":"We develop the statistical and algorithmic theory of inverse optimal transport (IOT) under the feature-parameterized cost C_theta(i,j) = -theta^T phi(i,j). The core technical contribution is the Sinkhorn linearization -- the implicit-function sensitivity of the entropic OT plan to the cost -- together with its spectral proxy, a formula that is spectrally exact yet geometrically transparent. The restricted Hessian on the tangent space satisfies the spectral sandwich (pi_min/epsilon) I = (pi_min/(a_max epsilon)) sqrt(lambda_min(Sigma)) that drives the entire theory. On this core we establish…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-08-13","reactions":{}}