Optimal transport between laws of random probability measures and the strict Monge problem
Pith reviewed 2026-05-09 17:02 UTC · model grok-4.3
The pith
For p>1 in strictly convex Banach spaces, optimal random couplings between laws of random probability measures are unique and induced by strict Monge maps.
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
Core claim
We consider an optimal transport problem between laws of random probability measures using a base cost function to build costs between measures and then between measures of measures. This admits a reformulation in terms of laws of random couplings. Optimal ones are characterized in terms of Kantorovich potentials. We introduce the strict Monge problem and give sufficient conditions under which its value coincides with the transport cost. For p>1 with cost equal to the distance to the power p in a strictly convex Banach space, sufficient conditions ensure the optimal random coupling is unique and induced by a solution of the strict Monge problem.
What carries the argument
Reformulation via laws of random couplings together with the strict Monge problem, enabling Kantorovich-potential characterization and equivalence to map-based solutions.
Load-bearing premise
Sufficient conditions on the Banach space, the random measures, and the base cost must hold to guarantee uniqueness of the optimal random coupling and its induction by a strict Monge map.
What would settle it
A concrete counterexample in a strictly convex Banach space for p>1 where the random measures meet the conditions but the optimal random coupling is neither unique nor induced by any single map.
read the original abstract
We consider an optimal transport problem between laws of random probability measures: given a base cost function, we build the associated OT cost between probability measures that in turn we use to define the OT cost between probability measures over probability measures. This setting admits a finer reformulation in terms of laws of random couplings, which retain more information than ordinary couplings. One of the main contributions of the paper is the characterization of the optimal ones in terms of Kantorovich potentials. Similarly, we also introduce the strict Monge problem, whose admissible competitors are more restrictive than in the usual Monge formulation. In this setting, we will give sufficient conditions under which the value of this problem is the same as the one considered above, in the spirit of the result by A. Pratelli. Then, for $p>1$, when the underlying cost is the distance to the power $p$ in a strictly convex Banach space, we will give sufficient conditions under which the optimal random coupling is unique and induced by a solution of the strict Monge problem, resembling the Brenier theorem.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript develops optimal transport between laws of random probability measures. Starting from a base cost, it constructs an OT cost first between probability measures and then between probability measures on the space of probability measures. The problem is reformulated in terms of laws of random couplings, which carry more information than ordinary couplings. The central contributions are a Kantorovich-potential characterization of optimal random couplings, the introduction of the strict Monge problem (with more restrictive competitors), sufficient conditions ensuring that the strict Monge problem attains the same value as the Kantorovich formulation (extending Pratelli), and, for p>1 with cost ||·||^p on a strictly convex Banach space, sufficient conditions under which the optimal random coupling is unique and induced by a map solving the strict Monge problem (Brenier-type result).
Significance. If the stated sufficient conditions hold and the proofs are complete, the work supplies a coherent extension of classical OT duality and uniqueness results to the random-measure setting. The Kantorovich characterization and the strict-Monge equivalence provide structural tools that could be useful in stochastic optimization and statistics; the Brenier-type uniqueness result is especially valuable because it reduces the problem to a map rather than a general coupling. The paper builds directly on standard Kantorovich duality and Pratelli’s theorem without evident circularity.
minor comments (3)
- §2 (or wherever the strict Monge problem is defined): an explicit side-by-side comparison of the admissible sets for the strict Monge problem versus the classical Monge problem would clarify the added restrictiveness and help readers assess the scope of the equivalence result.
- The sufficient conditions for the Brenier-type statement (p>1, strictly convex Banach space) are asserted in the abstract and presumably stated in the main theorems; a short remark on whether these conditions are sharp or admit simple counter-examples when violated would strengthen the presentation.
- Notation for random couplings and their laws is introduced early; a brief table or diagram contrasting ordinary couplings, random couplings, and their push-forwards would improve readability for readers less familiar with the random-measure setting.
Simulated Author's Rebuttal
We thank the referee for the careful reading and positive evaluation of our manuscript. The provided summary accurately captures the main contributions: the Kantorovich-potential characterization of optimal random couplings between laws of random measures, the introduction of the strict Monge problem with sufficient conditions for equivalence to the Kantorovich formulation (extending Pratelli), and the Brenier-type uniqueness result for p-power costs in strictly convex Banach spaces. We appreciate the noted significance for applications in stochastic optimization and statistics, and the observation that the work builds directly on classical OT duality without circularity. Since the report lists no specific major comments, we have no point-by-point responses to provide. We will make any minor revisions as needed in the revised version.
Circularity Check
No significant circularity
full rationale
The paper develops a theoretical framework for optimal transport between laws of random probability measures, characterizing optimal random couplings via Kantorovich potentials and establishing equivalence to the strict Monge problem under stated conditions on the space and cost. It extends standard Kantorovich duality and cites Pratelli’s result for the Monge equivalence, then provides sufficient conditions for uniqueness and strict Monge induction in the p>1 strictly convex Banach case, resembling Brenier’s theorem. All steps are deductive proofs from the problem definitions and duality, with no reduction of claims to self-definitions, fitted parameters renamed as predictions, or load-bearing self-citations; the derivation remains self-contained against external OT benchmarks.
Axiom & Free-Parameter Ledger
axioms (2)
- domain assumption Kantorovich duality and existence of potentials hold for the base optimal transport problem between probability measures.
- domain assumption The base cost function satisfies the conditions required for the p-power distance in strictly convex Banach spaces.
Reference graph
Works this paper leans on
-
[1]
Differentiability of Lipschitzian mappings between Banach spaces
isbn: 978-3-031-76833-0. doi: 10.1007/978-3-031-76834-7 . [AGS08] Luigi Ambrosio, Nicola Gigli, and Giuseppe Savar´ e. Gradient flows in metric spaces and in the space of probability measures . Second. Lectures in Mathematics ETH Z¨ urich. Birkh¨ auser Verlag, Basel, 2008, pp. x+334.isbn: 978-3-7643-8721-1. [Aro76] N. Aronszajn. “Differentiability of Lips...
-
[2]
doi: 10.1007/978-3-319-20828-2 . [Sch73] Laurent Schwartz. Radon measures on arbitrary topological spaces and cylindrical mea- sures. Vol. No. 6. Tata Institute of Fundamental Research Studies in Mathematics. Tata Institute of Fundamental Research, Bombay; by Oxford University Press, London, 1973, pp. xii+393. [Vau25] Cristophe Vauthier. “Variational Anal...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.