We study the statistical accuracy of Schrödinger bridge models based on an Ornstein-Uhlenbeck reference process. Given a known, compactly supported initial distribution and independent samples from an unknown target distribution, we estimate a right Schrödinger log-potential by minimizing empirical Kullback-Leibler (KL) risk over a prescribed function class. For bounded sub-Gaussian target densities, potentially with unbounded support, we establish a non-asymptotic high-probability oracle inequality for the KL-divergence between the target and estimated terminal marginals. Under additional two-sided Gaussian bounds and smoothness assumptions on the target log-density, we derive approximation guarantees for sparse ReLU networks and establish a KL convergence rate of O(n−2s/(2s+d)) for exact empirical risk minimization. This matches, up to logarithmic factors, the classical nonparametric KL-risk convergence exponent, while allowing target densities on Rd whose tails are controlled by Gaussian envelopes.
Figures & tables
Figure 1: Samples from the target distribution (green) and the estimated density (orange) in the experiments with Gaussian mixtures. From left to right, the panels show the standard and irregular targets with component covariance 0.1I2 , followed by the anisotropic target. Red markers indicate target centers.
Setup
Sliced Wasserstein
105MMD2
Energy distance
Standard
0.090±0.016
2.92±6.03
0.0025±0.0009
Irregular
0.098±0.040
2.36±12.59
0.0031±0.0028
Anisotropic
0.118±0.020
7.47±9.76
0.0040±0.0015
Table 1: Diagonal mixture fits with K=50 . Entries give the mean and standard deviation (divisor five) across five fresh evaluations of one fit per setup, each with 10,000 generated and 10,000 target samples.
Figure 2: Generated FFHQ faces from the bridge fitted to the child subset (left) and the female-labeled subset (right). Each panel shows all 25 samples from one draw.
We study nonparametric estimation of Schrödinger bridge (SB) drifts from i.i.d.\ data observed on a single time interval. Starting from the conditional-ratio form of the Schrödinger bridge time-series (SBTS) drift formula, we analyze a direct Nadaraya--Watson plug-in estimator built from kernelized numerator and denominator terms. Unlike recent SB analyses based on entropic-OT potentials, Sinkhorn iterations, or iterative bridge solvers, our approach works directly at the drift level and isolates statistical error from optimization, approximation, and discretization error. Under Hölder regularity, a marginal-density floor, and bounded support, we prove a uniform non-asymptotic bound for admissible bandwidth pairs, a pointwise CLT under genuine undersmoothing, and an adaptive bandwidth selector satisfying an oracle inequality. We also prove a pivot-local minimax lower bound which, through an explicit uniform pivot, yields a global minimax lower bound under transparent compatibility conditions; hence the adaptive selector is minimax-rate optimal up to logarithmic factors. Synthetic experiments provide theorem-targeted diagnostics for finite-sample scaling, Gaussian approximation, and adaptive behavior.
Othmane Mazhar, Huyên Pham
LPSM, Sorbonne Université and Université Paris Cité · École Polytechnique, CMAP
The Schrödinger bridge owes its computational power to a single structural fact: by Girsanov's theorem the controlled problem is a Kullback--Leibler (KL) projection onto a fixed reference measure, solvable by alternating projections. This letter shows that the fact does not survive risk sensitivity. When the expected path cost is replaced by the entropic risk measure and both endpoint marginals are kept as hard constraints, the resulting fixed-point bridge value Jθ (the soft-problem value at the multiplier that enforces the terminal constraint) admits no representation as a constrained KL minimum against any fixed path-space reference with a regular endpoint law (a class strictly larger than the uniformly elliptic diffusion references: no Markov property is required), even allowing an additive normalisation depending on the initial marginal. Moreover, no single reference generates the one-parameter family in the risk parameter. The obstruction is computed in closed form: the Gaussian bridge value violates, by exactly θ/2, a heat equation that any Gaussian smoothing of a fixed endpoint density must obey. In place of the projection, the theory rests on a terminal-multiplier fixed point and an asymmetric factorisation penalising the score energy of the backward factor.
We propose a closed-form spectral framework for relative log-density estimation in linearly parameterized probabilistic models, including unnormalized and conditional models. This is achieved by representing the Kullback-Leibler (KL) divergence as an integral of weighted chi-squared divergences, converting KL estimation into a family of least-squares problems. We derive an explicit spectral formula based only on first- and second-order feature moments, yielding closed-form estimators of both divergences and log-density potentials for fixed features. The framework extends to a broad class of f-divergences and can be combined with kernelization or feature learning with neural networks (with then only a closed form for the last layer). We prove convergence guarantees for the resulting estimators and empirically compare them on synthetic data with optimization-based variational formulations, including logistic and softmax regression for normalized conditional models.
Francis Bach
SIERRA · Inria - Ecole Normale Supérieure, PSL Research University