pith. sign in

arxiv: 2505.09232 · v3 · submitted 2025-05-14 · 🧮 math.AP · math.OC

Absence of loops for the Wasserstein-mathcal{H}¹ problem: the concentration/blow-up argument

Jo{\~a}o Miguel Machado (LMCRC) This is my paper

Pith reviewed 2026-05-22 15:48 UTC · model grok-4.3

classification 🧮 math.AP math.OC
keywords Wasserstein-H1 problemtree minimizersabsence of loopsconcentration argumentblow-up argumentoptimal transportDirac measuresbounded density
0
0 comments X

The pith

Minimizers of the Wasserstein-H^1 problem are trees when the target measure is a finite sum of Dirac masses or has bounded density.

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

The paper establishes that solutions to the Wasserstein-H^1 optimal transport problem take the form of trees, without any loops, in two specific situations. This occurs when the target measure consists of finitely many Dirac point masses or when the target has a density that is bounded from above. The proof relies on a concentration and blow-up technique to show that assuming a loop leads to a contradiction with the minimality of the configuration. Readers may care about this because tree structures are simpler to understand and could lead to better ways of computing or approximating these transport plans.

Core claim

The central discovery is that minimizers of the Wasserstein-mathscr{H}^1 problem are trees in the cases where the target measure is a sum of finitely many Dirac masses or when it has a bounded density. This is established through a concentration/blow-up argument that produces a contradiction if a loop is present in the minimizer.

What carries the argument

The concentration/blow-up argument, which assumes the existence of a loop and then concentrates or rescales to violate optimality.

If this is right

  • Minimizers have acyclic support.
  • The tree property holds for finite atomic targets.
  • The tree property holds for targets with bounded density.
  • The argument excludes cycles in the geometric structure of the solution.

Where Pith is reading between the lines

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

  • Numerical algorithms could be designed to optimize only over tree configurations.
  • The result links to problems in geometric optimization like the Steiner tree problem.
  • Extensions might consider targets with densities that are unbounded but still integrable.

Load-bearing premise

The existence of minimizers for the Wasserstein-H^1 problem is guaranteed for the measures considered.

What would settle it

Exhibiting a minimizer that contains a closed loop for a target measure consisting of two Dirac masses would disprove the claim.

Figures

Figures reproduced from arXiv: 2505.09232 by Jo{\~a}o Miguel Machado (LMCRC).

Figure 1
Figure 1. Figure 1: Heuristic proof of existence of an optimal shape for problem (PΛ). If a solution has an excess part, represented in the figure by a measure having a density along Σ and a Dirac mass, it must be formed through projections onto Σ. But then it is better to send the excess mass that is being projected to small segments in the direction of the projection [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Argument for absence of loops for (P Λ). As in the proof of existence, we begin by showing that loops are formed through projections and later use this information to construct a better competitor. 1.1. Contributions and the concentration/blow-up argument. As previously stated, in this work we show that the support of minimizers of (P Λ) are trees in three cases Case 1: if ̺0 is a convex combination of Dir… view at source ↗
Figure 3
Figure 3. Figure 3: Construction of a better competitor in Theorem 3.6. On the right, the partition of the space into sections. For sections i, i′ such that ¯θi , ¯θi ′ > 0 we add a segment in their direction. For ¯θj , ¯θj ′ = 0 we construct a Dirac mass. On the cases of positive density we have a gain of order ε 2 in transportation cost, for zero density we lose o(ε 2 ). On the left the transportation strategy of each secti… view at source ↗
read the original abstract

In the present work we prove that minimizers of the Wasserstein-$\mathscr{H}^1$ problem, introduced recently by Chambolle et. al., are trees in two cases: when the target measure is a sum of finitely many Dirac masses or when it has a bounded density.

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

1 major / 2 minor

Summary. The manuscript proves that minimizers of the Wasserstein-ℋ¹ problem are trees (i.e., their supports contain no closed loops) in two cases: when the target measure is a finite sum of Dirac masses or when it has bounded density. The central argument is a concentration/blow-up construction: the presence of a loop allows rescaling to produce a strictly lower-energy competitor in the limit, contradicting minimality. The proof invokes existence and lower-semicontinuity results from Chambolle et al. as external input together with standard compactness properties in the Wasserstein space.

Significance. If the result holds, it supplies a clean structural characterization of minimizers for the Wasserstein-ℋ¹ functional, confirming they are acyclic for the two classes of target measures. This is useful for the field because it restricts the possible geometries of optimal configurations and may facilitate further analysis or numerics. The paper delivers a direct, parameter-free contradiction argument via blow-up; this is a strength that makes the claim falsifiable on simple test cases with finitely many Diracs.

major comments (1)
  1. [§4] §4 (Dirac-mass case): the blow-up construction assumes the loop lies at positive distance from all atoms of the target measure. The argument must explicitly rule out loops that touch or connect to a Dirac point, because the transport cost to that atom could change under rescaling and potentially invalidate the strict energy decrease.
minor comments (2)
  1. Notation for the functional is inconsistent (ℋ¹ vs. ℋ¹); adopt a single symbol throughout the text and in the title.
  2. [Introduction] The introduction would benefit from a short paragraph contrasting the present blow-up method with existing loop-removal techniques in branched transport or irrigation problems.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the positive assessment of the manuscript and for the recommendation of minor revision. The single major comment is well-taken and points to a case that requires explicit treatment in the Dirac-mass argument. We address it below and will incorporate the necessary clarification and case distinction into the revised version of Section 4.

read point-by-point responses

  1. Referee: [§4] §4 (Dirac-mass case): the blow-up construction assumes the loop lies at positive distance from all atoms of the target measure. The argument must explicitly rule out loops that touch or connect to a Dirac point, because the transport cost to that atom could change under rescaling and potentially invalidate the strict energy decrease.

    Authors: We agree that the current write-up of the blow-up argument in the finite-Dirac case implicitly assumes the loop lies at positive distance from every atom. To close this gap we will add a separate case analysis. When a loop touches or connects to a Dirac atom, we perform the concentration at a point of the loop that is not the atom itself and then adjust the optimal transport plan by cutting the loop at the connection point and reassigning the infinitesimal mass to the atom along a shorter path. Because the atom is a point mass, this local modification decreases the total H^1 length while preserving the marginals and the Wasserstein cost up to a higher-order term that vanishes in the blow-up limit. The resulting competitor therefore yields a strict energy decrease, again contradicting minimality. The revised Section 4 will contain this case distinction together with the corresponding estimates. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation is self-contained

full rationale

The manuscript supplies a concentration/blow-up argument showing that any closed loop in a candidate minimizer can be rescaled to a strictly lower-energy competitor, contradicting minimality for the two classes of target measures. This rests on the existence theory from Chambolle et al. treated as external input together with standard lower-semicontinuity and compactness of the Wasserstein-H^1 functional; no step reduces by construction to a fitted parameter, self-definition, or self-citation chain. The central structural claim therefore remains independent of the paper's own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper rests on existence and regularity theory from the cited Chambolle et al. work; no new free parameters or invented entities are introduced in the abstract statement.

axioms (1)
  • domain assumption Existence of minimizers for the Wasserstein-H^1 functional holds for the target measures considered.
    Invoked implicitly to state that minimizers exist and can be analyzed for their support structure.

pith-pipeline@v0.9.0 · 5570 in / 1181 out tokens · 24085 ms · 2026-05-22T15:48:38.616872+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]

    Lectures on optimal transport

    Luigi Ambrosio, Elia Brué, and Daniele Semola. Lectures on optimal transport . Springer, 2021

    work page 2021

  2. [2]

    Functions of bounded variation and free discontinuity prob lems

    Luigi Ambrosio, Nicola Fusco, and Diego Pallara. Functions of bounded variation and free discontinuity prob lems. Courier Corporation, 2000

    work page 2000

  3. [3]

    On the history of the euclidean steiner tree problem

    Marcus Brazil, Ronald L Graham, Doreen A Thomas, and Mart in Zachariasen. On the history of the euclidean steiner tree problem. Archive for history of exact sciences , 68(3):327–354, 2014

    work page 2014

  4. [4]

    Optimal transpo rtation networks as free dirichlet regions for the monge- kantorovich problem

    Giuseppe Buttazzo and Eugene Stepanov. Optimal transpo rtation networks as free dirichlet regions for the monge- kantorovich problem. Annali della Scuola Normale Superiore di Pisa-Classe di Scie nze, 2(4):631–678, 2003

    work page 2003

  5. [5]

    One dimensional approximation of measures in W asser- stein distance

    Antonin Chambolle, Vincent Duval, and João Miguel Macha do. One dimensional approximation of measures in W asser- stein distance. Journal de l’École Polytechnique-Mathématiques , 12:101–145, 2025

    work page 2025

  6. [6]

    Lecture notes on rectifiable sets, den sities, and tangent measures

    Camillo De Lellis. Lecture notes on rectifiable sets, den sities, and tangent measures. Preprint, 23, 2006

    work page 2006

  7. [7]

    A presentation of the average distanc e minimizing problem

    Antonie Lemenant. A presentation of the average distanc e minimizing problem. Journal of Mathematical Sciences , 181(6), 2012

    work page 2012

  8. [8]

    Phase-field approximation for one- dimensional shape optimization problems

    João Miguel Machado. Phase-field approximation for one- dimensional shape optimization problems. SIAM Journal on Mathematical Analysis, 57(4):3488–3523, 2025

    work page 2025

  9. [9]

    Sets of finite perimeter and geometric variational problems : an introduction to Geometric Measure Theory

    Francesco Maggi. Sets of finite perimeter and geometric variational problems : an introduction to Geometric Measure Theory. Number 135. Cambridge University Press, 2012

    work page 2012

  10. [10]

    Γ-convergence for the irrigation problem

    Sunra JN Mosconi, Paolo Tilli, et al. Γ-convergence for the irrigation problem. J. Convex Anal , 12(1):145–158, 2005

    work page 2005

  11. [11]

    S tructure of average distance minimizers in general dimensi ons

    Lucas O’Brien, Forest Kobayashi, and Young-Heon Kim. S tructure of average distance minimizers in general dimensi ons. arXiv preprint arXiv:2503.23256 , 2025

    work page arXiv 2025

  12. [12]

    Qualitative pro perties of maximum distance minimizers and average distanc e minimizers in rn

    Emanuele Paolini and Eugene Stepanov. Qualitative pro perties of maximum distance minimizers and average distanc e minimizers in rn. Journal of Mathematical Sciences , 122(3):3290–3309, 2004

    work page 2004

  13. [13]

    Existence and re gularity results for the steiner problem

    Emanuele Paolini and Eugene Stepanov. Existence and re gularity results for the steiner problem. Calculus of Variations and Partial Differential Equations , 46(3):837–860, 2013

    work page 2013

  14. [14]

    Variational analysis, volume 317

    R Tyrrell Rockafellar and Roger J-B W ets. Variational analysis, volume 317. Springer Science & Business Media, 2009

    work page 2009

  15. [15]

    Santambrogio

    F. Santambrogio. Optimal transport for applied mathem aticians. Birkäuser, NY , 55(58-63):94, 2015

    work page 2015

  16. [16]

    Blow-up of optim al sets in the irrigation problem

    Filippo Santambrogio and Paolo Tilli. Blow-up of optim al sets in the irrigation problem. The Journal of Geometric Analysis, 15:343–362, 2005

    work page 2005

  17. [17]

    Optimal transport: old and new , volume 338

    Cédric Villani. Optimal transport: old and new , volume 338. Springer, 2009. Lagrange Mathematical and Computing Center, 103 rue de Gren elle, Paris, 75007 Email address : joao-miguel.machado@ceremade.dauphine.fr

    work page 2009