Uniform Statistical Convergence of Empirical Sinkhorn Potentials with Exponential and Polynomial Dependence on the Regularization Parameter
Abstract
We study the empirical Sinkhorn estimator of the entropic optimal transport potentials under the uniform loss. Since the potentials are only unique up to additive constants, we measure the error using the quotient supremum norm, defined as d∞([u],[v]) = infa∈ℝ|u−v−a|∞. For a fixed regularization parameter ε>0, we establish a non-asymptotic statistical rate of n−1/2. This is achieved by combining the Birkhoff-Hopf contraction theorem with entropy bounds on normalized kernel sections. However, the constant in this bound grows exponentially with 1/ε. To improve this, we isolate geometric conditions under which the empirical estimator maintains the n−1/2 rate but features polynomial dependence on 1/ε. The key requirement is a polynomial residual-stability estimate for the population Sinkhorn map. We provide sufficient criteria for this, including a polynomial contraction property and a local inverse estimate. Furthermore, we introduce two rigorously verifiable model classes an ε-weak residual-interaction class obtained after separable centering and another based on connected tight-edge graphs for fixed discrete costs where the polynomial rate is guaranteed without relying on abstract resolvent assumptions. Finally, we establish matching minimax lower bounds demonstrating that the ε n−1/2 rate cannot be uniformly improved in the bounded-interaction regime.
Related
- Finite-Data Error Bounds for Approximating the Koopman Operator: Sampling Measures, Super-Polynomial Convergence and Regularization
- Combining Exponential and Polynomial Convergence Rates for Accelerated Extremum Seeking: A High-Order Lie Bracket Approach
- Stability of Kantorovich potentials via heat kernel regularization
- Largest-Angle Bisection and Statistical Regularization of Triangle Shapes
- On the exponential ergodicity of the McKean-Vlasov SDE depending on a polynomial interaction
BibTeX
@misc{belomestny2026uniform,
title = {{Uniform Statistical Convergence of Empirical Sinkhorn Potentials with Exponential and Polynomial Dependence on the Regularization Parameter}},
author = {Denis Belomestny},
howpublished = {arXiv},
year = {2026},
month = aug,
doi = {10.48550/arxiv.2608.29152},
eprint = {2608.29152},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2608.29152},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Uniform Statistical Convergence of Empirical Sinkhorn Potentials with Exponential and Polynomial Dependence on the Regularization Parameter
AU - Denis Belomestny
PY - 2026
DA - 2026/08/29
PB - arXiv
DO - 10.48550/arxiv.2608.29152
UR - https://arxiv.org/abs/2608.29152
AB - We study the empirical Sinkhorn estimator of the entropic optimal transport potentials under the uniform loss. Since the potentials are only unique up to additive constants, we measure the error using the quotient supremum norm, defined as d_∞([u],[v]) = inf_a∈ℝ|u−v−a|_∞. For a fixed regularization parameter ε>0, we establish a non-asymptotic statistical rate of n^−1/2. This is achieved by combining the Birkhoff-Hopf contraction theorem with entropy bounds on normalized kernel sections. However, the constant in this bound grows exponentially with 1/ε. To improve this, we isolate geometric conditions under which the empirical estimator maintains the n^−1/2 rate but features polynomial dependence on 1/ε. The key requirement is a polynomial residual-stability estimate for the population Sinkhorn map. We provide sufficient criteria for this, including a polynomial contraction property and a local inverse estimate. Furthermore, we introduce two rigorously verifiable model classes an ε-weak residual-interaction class obtained after separable centering and another based on connected tight-edge graphs for fixed discrete costs where the polynomial rate is guaranteed without relying on abstract resolvent assumptions. Finally, we establish matching minimax lower bounds demonstrating that the ε n^−1/2 rate cannot be uniformly improved in the bounded-interaction regime.
ER -