pith. sign in

arxiv: 1907.07059 · v1 · pith:IXNT3FQJnew · submitted 2019-07-16 · 🧮 math.PR

A note on duality theorems in mass transportation

Pith reviewed 2026-05-24 20:43 UTC · model grok-4.3

classification 🧮 math.PR
keywords duality theoremsmass transportationMonge-Kantorovich problemoptimal transportprobability spacescost functionsemicontinuous functionsperfect measures
0
0 comments X

The pith

Duality theorems for the Monge-Kantorovich problem hold in any probability spaces when the cost is bounded between sums of integrable functions, without needing perfect measures.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper proves duality equalities for the infimum α(c) and supremum α*(c) of integrals of a measurable cost c over all couplings of two probability measures μ and ν. These equalities are established in a fully abstract setting once c satisfies a sandwich bound by integrable functions from each space. The results cover cases where the measures need not be perfect and extend to upper or lower semicontinuous costs when the spaces are metric and one measure is separable. This matters because it removes a common technical restriction that fails in many natural examples and yields explicit duality statements for continuous or bounded costs as well.

Core claim

Under the condition that there exist f1, f2 in L1(μ) and g1, g2 in L1(ν) with f1 + g1 ≤ c ≤ f2 + g2, the quantities α(c) and α*(c) satisfy duality equalities with their natural dual formulations, and this holds for arbitrary probability spaces without any perfectness assumption on μ or ν. When the underlying spaces are metric and μ is separable, duality for α(c) holds whenever c is upper semicontinuous, duality for α*(c) holds whenever c is lower semicontinuous, and duality holds simultaneously for both when the maps x ↦ c(x,y) and y ↦ c(x,y) are continuous (or when c is bounded with one family of sections continuous). These statements improve earlier results whenever the cardinalities of X,

What carries the argument

The sandwich condition f1 + g1 ≤ c ≤ f2 + g2 with fi integrable with respect to μ and gi integrable with respect to ν, which supplies the uniform integrability needed to equate the primal infima and suprema with their dual expressions.

If this is right

  • When the spaces are metric and μ is separable, duality for α(c) holds if c is upper semicontinuous.
  • Duality for α*(c) holds if c is lower semicontinuous under the same metric and separability assumptions.
  • Duality holds simultaneously for both α(c) and α*(c) if the sections x ↦ c(x,y) and y ↦ c(x,y) are continuous.
  • Duality holds for both if c is bounded and one family of sections is continuous.
  • The stated duality statements improve earlier theorems whenever the underlying sets have cardinality at most the continuum.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The sandwich condition may be checkable in applications where costs arise from distances or utilities that grow at most linearly.
  • Results of this type could be used to justify passage to the limit in transport problems on non-separable spaces once a dominating integrable bound is verified.
  • One could search for minimal weakenings of the sandwich that still guarantee duality, for instance by replacing global integrability with local conditions.
  • The improvement for sets of cardinality at most the continuum suggests checking whether the same statements remain valid for larger cardinals under additional set-theoretic assumptions.

Load-bearing premise

There must exist integrable functions f1, f2 on the first space and g1, g2 on the second space such that the cost c lies between their sums.

What would settle it

A concrete triple of probability spaces, marginals μ and ν, and cost c obeying the sandwich bound for which α(c) fails to equal its dual expression would falsify the claimed duality.

read the original abstract

The duality theory of the Monge-Kantorovich transport problem is investigated in an abstract measure theoretic framework. Let $(\mathcal{X},\mathcal{F},\mu)$ and $(\mathcal{Y},\mathcal{G},\nu)$ be any probability spaces and $c:\mathcal{X}\times\mathcal{Y}\rightarrow\mathbb{R}$ a measurable cost function such that $f_1+g_1\le c\le f_2+g_2$ for some $f_1,\,f_2\in L_1(\mu)$ and $g_1,\,g_2\in L_1(\nu)$. Define $\alpha(c)=\inf_P\int c\,dP$ and $\alpha^*(c)=\sup_P\int c\,dP$, where $\inf$ and $\sup$ are over the probabilities $P$ on $\mathcal{F}\otimes\mathcal{G}$ with marginals $\mu$ and $\nu$. Some duality theorems for $\alpha(c)$ and $\alpha^*(c)$, not requiring $\mu$ or $\nu$ to be perfect, are proved. As an example, suppose $\mathcal{X}$ and $\mathcal{Y}$ are metric spaces and $\mu$ is separable. Then, duality holds for $\alpha(c)$ (for $\alpha^*(c)$) provided $c$ is upper-semicontinuous (lower-semicontinuous). Moreover, duality holds for both $\alpha(c)$ and $\alpha^*(c)$ if the maps $x\mapsto c(x,y)$ and $y\mapsto c(x,y)$ are continuous, or if $c$ is bounded and $x\mapsto c(x,y)$ is continuous. This improves the existing results in \cite{RR1995} if $c$ satisfies the quoted conditions and the cardinalities of $\mathcal{X}$ and $\mathcal{Y}$ do not exceed the continuum.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

0 major / 3 minor

Summary. The manuscript proves duality theorems for the Monge-Kantorovich problems α(c) = inf ∫ c dP and α*(c) = sup ∫ c dP (inf/sup over couplings P with fixed marginals μ, ν) in an abstract measure-theoretic setting. The key assumption is the sandwich condition f1 + g1 ≤ c ≤ f2 + g2 with fi ∈ L1(μ) and gi ∈ L1(ν); under this condition duality holds without requiring μ or ν to be perfect. Applications are given for upper/lower semicontinuous costs when the spaces are metric and μ is separable, and for continuous or bounded costs, improving on results from RR1995 when |X| and |Y| ≤ continuum.

Significance. If the proofs hold, the work is significant because it removes the perfectness assumption that is standard but restrictive in the transport literature, while the explicit sandwich condition makes the integrals well-defined and finite. The concrete semicontinuity and continuity corollaries supply usable sufficient conditions on metric spaces with separable measures.

minor comments (3)
  1. [Abstract / Introduction] The abstract states that duality theorems are proved but does not display the precise dual expressions (e.g., the form of the sup over f + g ≤ c). Adding the exact statement of the duality equalities in the introduction or as Theorem 2.1 would make the contribution immediately clear.
  2. [Introduction (last paragraph)] The improvement over RR1995 is stated to hold when cardinalities do not exceed the continuum; the manuscript should indicate whether this cardinality restriction is essential to the argument or merely an artifact of the comparison.
  3. [References] The reference RR1995 appears only as a citation; the bibliography entry should be supplied in full so readers can locate the precise statements being improved upon.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary and recommendation of minor revision. The report contains no specific major comments to address point by point.

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper proves duality theorems for α(c) and α*(c) directly from the sandwich condition f1+g1 ≤ c ≤ f2+g2 with f1,f2 ∈ L1(μ) and g1,g2 ∈ L1(ν), using standard measure-theoretic arguments on arbitrary probability spaces. No step reduces by construction to its inputs, no fitted parameter is relabeled as a prediction, and the improvement over RR1995 is presented as a consequence of the new setting rather than a self-citation chain. The central claims remain independent of the cited prior work and do not invoke uniqueness theorems or ansatzes smuggled via self-reference.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The paper relies on standard axioms of measure theory and probability; no free parameters or new entities introduced.

axioms (2)
  • standard math The spaces are probability spaces with sigma-algebras and measures
    Basic setup for the transport problem.
  • domain assumption The cost function is measurable
    Required for the integrals to be defined.

pith-pipeline@v0.9.0 · 5855 in / 1205 out tokens · 27337 ms · 2026-05-24T20:43:17.644184+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

  1. [1]

    (2008) Gradient flows , Second Edition, Birkhauser, Basel

    Ambrosio L., Gigli N., Savare’ G. (2008) Gradient flows , Second Edition, Birkhauser, Basel

  2. [2]

    (1974) Operator algebras and invariant subsp aces, Ann

    Arveson W. (1974) Operator algebras and invariant subsp aces, Ann. of Math. , 100, 433-533

  3. [3]

    (2011) Duality for Borel measurable cost functions, Trans

    Beiglbock M., Schachermayer W. (2011) Duality for Borel measurable cost functions, Trans. Amer. Math. Soc. , 363, 4203-4224

  4. [4]

    (2012) A gene ral duality theorem for the Monge-Kantorovich transport problem, Studia Math

    Beiglbock M., Leonard C., Schachermayer W. (2012) A gene ral duality theorem for the Monge-Kantorovich transport problem, Studia Math. , 209, 151-167

  5. [5]

    (2015) Two versions of the f undamental theorem of asset pricing, Electronic J

    Berti P., Pratelli L., Rigo P. (2015) Two versions of the f undamental theorem of asset pricing, Electronic J. Probab. , 20, 1-21

  6. [6]

    (2015) Equ ivalent or absolutely continuous prob- ability measures with given marginals, Dependence Modeling, 3, 47-58

    Berti P., Pratelli L., Rigo P., Spizzichino F. (2015) Equ ivalent or absolutely continuous prob- ability measures with given marginals, Dependence Modeling, 3, 47-58

  7. [7]

    (1996) On a measure-theoretic prob lem of Arveson, Proc

    Haydon R., Shulman V. (1996) On a measure-theoretic prob lem of Arveson, Proc. Amer. Math. Soc. , 124, 497-503

  8. [8]

    (1942) On the translocation of masses, C

    Kantorovich L. (1942) On the translocation of masses, C. R. (Doklady) Acad. Sci. URSS (N.S.), 37, 199-201. 12 PIETRO RIGO

  9. [9]

    (1984) Duality theorems for marginal prob lems, Z

    Kellerer H.G. (1984) Duality theorems for marginal prob lems, Z. Wahrscheinlichkeitstheorie Verw. Geb. , 67, 399-432

  10. [10]

    (1981) On perfect measures, Trans

    Koumoullis G. (1981) On perfect measures, Trans. Amer. Math. Soc. , 264, 521-537

  11. [11]

    (2015) Extremal dependence concep ts, Statist

    Puccetti G., W ang R. (2015) Extremal dependence concep ts, Statist. Science , 30, 485-517

  12. [12]

    (2019) Centers of p robability measures without the mean, J

    Puccetti G., Rigo P., W ang B., W ang R. (2019) Centers of p robability measures without the mean, J. Theore. Probab. , 32, 1482-1501

  13. [13]

    (1998) Mass transportation problems, Volume I: Theory , Springer, New York

    Rachev S.T., Ruschendorf L. (1998) Mass transportation problems, Volume I: Theory , Springer, New York

  14. [14]

    (1995) A general duali ty theorem for marginal problems, Prob

    Ramachandran D., Ruschendorf L. (1995) A general duali ty theorem for marginal problems, Prob. Theo. Relat. Fields , 101, 311-319

  15. [15]

    (2000) On the Monge-Ka ntorovitch duality theorem, The- ory Probab

    Ramachandran D., Ruschendorf L. (2000) On the Monge-Ka ntorovitch duality theorem, The- ory Probab. Appl. , 45, 350-356

  16. [16]

    (2013) Mathematical Risk Analysis: Dependence, Risk Bounds, Opti mal Allocations and Portfolios , Springer, Heidelberg

    Ruschendorf L. (2013) Mathematical Risk Analysis: Dependence, Risk Bounds, Opti mal Allocations and Portfolios , Springer, Heidelberg

  17. [17]

    F. Casorati

    Villani C. (2009) Optimal transport, old and new , Springer, New York. Pietro Rigo, Dipartimento di Matematica “F. Casorati”, Univ ersita’ di Pavia, via Ferrata 1, 27100 Pavia, Italy E-mail address : pietro.rigo@unipv.it