REVIEW 1 major objections 4 minor 30 references
Folded optimal transport and its application to separable quantum optimal transport
T0 review · 1 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash
Pith's one-line read Folded optimal transport extends any distance on the extreme points of a convex set to a distance on the whole set, yielding a genuine separable quantum Wasserstein distance on density matrices.
desk verdict Solid construction, real unification, one overclaimed remark about Fubini-Study and arbitrary norms that needs correcting but does not sink the paper. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The key object is the folded Wasserstein (pseudo-)distance. First, D̂_p(x,y) = inf { W_p(µ,ν) : µ represents x, ν represents y } glues the ordinary Wasserstein distance on P(E) along the Choquet equivalence classes of representing measures; this yields a semi-distance (symmetric, separating, but not always subadditive). Then D_p(x,y) is its chain closure, the infimum of Σ D̂_p(z_i,z_i+1) over all finite chains from x to y, which enforces the triangle inequality. The load-bearing identity is the Choquet identification C ≃ P(E)/∼, which turns a convex set into a quotient of a Wasserstein space; the condition d(x,y) ≥ ||x−y|| on the boundary is what forces D_p to separate points and to upper-bo
What would settle it
Check the paper's claim that the Fubini-Study metric satisfies d ≥ ||x−y|| for any norm on B(H): for two orthogonal pure states in dimension 2, d_FS(P,Q) = π/2 but ||P−Q||_1 = 2, so the inequality fails for the trace norm. This shows the hypothesis of Theorem 3 must be verified for each chosen norm and that the remark preceding it is not correct.
Extended reading notes
Core claim
The central claim is that the folded Wasserstein distance D_p — defined on a compact convex set C by first lifting a distance d on the extreme boundary E to the Wasserstein distance W_p on probability measures over E, then minimizing W_p among measures representing the same points, and finally taking the shortest-chain closure — is an honest distance when the boundary distance dominates the ambient norm (d(x,y) ≥ ||x−y||). Under that hypothesis, (C,D_p) is compact, Polish, continuous with respect to the norm topology when the relative interior is nonempty, and geodesic whenever (E,d) is geodesic and p>1. In the quantum case C = S_+^1 (density matrices), E = PH (pure states), the construction
Load-bearing premise
The entire construction hinges on the boundary distance dominating the ambient norm: d(x,y) ≥ ||x−y|| for all extreme points x,y; if this fails, the paper's proof that D_p separates points and induces the natural topology collapses, and D_p may degenerate.
Editorial extensions
If this is right
- Under the boundary-dominance condition, the space of density matrices (S_+^1, D_p) becomes a compact Polish geodesic metric space for p>1 when the pure-state space is geodesic, so quantum states acquire a bona fide metric geometry inherited purely from a distance on pure states.
- The Beatty–Stilck França semi-distance is exactly the folded Kantorovich semi-distance D̂_p; the paper improves the optimal transport plan to at most 2 dimH − 1 atoms and shows D̂_p always separates points.
- The Golse–Paul semiclassical cost is a folded Kantorovich cost, so semiclassical comparisons between quantum and classical states fit into the same optimal-transport framework and extend naturally to arbitrary Radon measures rather than only densities.
- Because the construction recovers classical Wasserstein when the convex set is a simplex, classical results about Wasserstein spaces transfer to general convex state spaces via the folding recipe.
- The folded distance D_p sub-extends the boundary distance d in general, and extends it exactly when d is a norm; this clarifies when a quantum Wasserstein distance can preserve pure-state distances literally.
Reading between the lines
- The same folding recipe could be applied to other convex state spaces with well-understood extreme boundaries, such as Gaussian states or fermionic density matrices; any distance on the pure/physical boundary that dominates the ambient norm would automatically produce a valid Wasserstein-type metric.
- The identification of the Golse–Paul cost suggests that quantitative semiclassical limits (Wigner measures, mean-field limits) could be studied by importing standard optimal-transport stability results through the folded cost.
- The 2 dimH − 1 atom bound hints that the quantum folded transportation problem is a linear program with a dimension-dependent rank, possibly making it computationally competitive for small systems; one could test whether the bound is tight.
- Because D_p is defined by a chain closure, it inherits a dynamical/geodesic interpretation when the boundary is geodesic; this may open a path to gradient-flow formulations for quantum entropies under the folded metric, in analogy with classical Wasserstein gradient flows — a speculation not made in the paper.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper introduces 'folded optimal transport', a two-step procedure (lift to the Wasserstein space over the extreme boundary, then quotient by the barycenter equivalence relation) that extends a distance or cost defined on the extreme boundary E of a compact convex set C to the whole of C. The main general result, Theorem 1, asserts that under the lower-bound condition d(x,y) ≥ ||x−y|| on E, the folded Wasserstein pseudo-distance D_p is an actual distance, is compatible with the natural topology, is compact/Polish when C is finite-dimensional or appropriate compactness holds, is geodesic when d is geodesic and p>1, and dominates the ambient norm. The paper then specializes to finite-dimensional quantum state space S_+^1 with extreme boundary PH, proving Theorem 3 and Corollary 1 for the Frobenius and Fubini–Study pure-state distances. It identifies the folded Kantorovich semi-distance with the Beatty–Stilck França semi-distance, improves the number of atoms in optimal representing plans to 2 dim H − 1, proves separation and continuity of the latter semi-distance without the Hölder-continuity assumption raised in [4], and recasts the Golse–Paul semiclassical cost as a folded Kantorovich cost.
Significance. If the claims hold, the framework is a valuable unification: standard Wasserstein distances, a class of separable quantum Wasserstein distances, and the Golse–Paul semiclassical cost all arise from one construction. The paper is careful and mostly self-contained: the LP representation of the folded Kantorovich semi-distance, the existence and atomicity of optimal plans via Winkler's theorem, and the continuity/topology arguments are detailed and credible. There are no fitted parameters, the identification with the Beatty–Stilck França semi-distance is proven rather than assumed, and the atom-counting improvement is a concrete advance. The main reservation is a scope error in the quantum section: the claim that the Fubini–Study metric satisfies the key domination assumption for every norm is false, and the advertised applicability of Theorem 3 must be corrected. I do not see this as a fatal flaw, because the theorem itself is conditional and the proof is sound; the fix is local but affects the statement of the main quantum result.
major comments (1)
- [Section 3.2, paragraph before Theorem 3] The text states that 'the required assumption (31) is satisfied by the Fubini-Study metric and any norm on B(H)'. This is incorrect. For the Frobenius norm, write |⟨ψ|φ⟩| = cos θ; then d_FS(P_ψ,P_φ) = θ while ||P_ψ − P_φ||_Fr = √2 sin θ. For small θ we have √2 sin θ > θ, so d_FS(P_ψ,P_φ) < ||P_ψ − P_φ||_Fr. The same failure occurs for the trace norm. Thus Theorem 3 cannot be invoked with d_FS and the standard Frobenius or trace norm. The correct statement is that for any fixed norm there is an equivalent norm — e.g. 2^{-1/2}||·||_Fr for d_FS — for which (31) holds; Corollary 1(ii) implicitly uses this rescaling. Please correct the parenthetical and make the reference norm explicit at every point where (31) is applied.
minor comments (4)
- [Section 1] There are unresolved figure placeholders 'Figures??and??' in the paragraph around equation (3).
- [Corollary 1 proof] In the first paragraph of the proof, the phrase 'Therefore d_FS obviously induces the natural topology and upper-bounds ||·||_Fr' appears to be a typo: it should refer to d_Fr, not d_FS. The subsequent paragraph gives the correct d_FS inequality.
- [Throughout] Spelling is inconsistent: 'Fubini-Study' is misspelled as 'Fubini–Stud' in a few places, and the author name is given as both 'Beatty–Stilck França' and 'Beatty–França'. Please standardize.
- [Proposition 9] The word 'representant' should be 'representing' in 'optimal representant transport plan'.
Circularity Check
No circularity: the folded Wasserstein construction is derived from standard optimal transport and Choquet theory, with the Beatty–França identification proved rather than assumed.
full rationale
The paper's derivation is self-contained. The folded Wasserstein distance is defined in (7)–(8) as the quotient pseudo-distance obtained from the standard Wasserstein distance W_p on P(E) modulo the Choquet equivalence relation (5)–(6); it is not defined in terms of the metric conclusions it later derives. Theorem 1's distance and topology claims follow from standard facts about W_p metrizing weak-* convergence, compactness of (P(E), W_p), closedness of representing-coupling sets (Lemma 2), and Lemma 1's duality argument proving D_p ≥ ‖·‖ when d ≥ ‖·‖. There are no fitted parameters and no prediction obtained by renaming a fitted input. In the quantum section, the identification of the Beatty–França semi-distance with the folded Kantorovich semi-distance is established in Proposition 7 by showing the two linear programs coincide and that a finitely supported optimal plan exists (Proposition 9); it is not assumed. The geometric properties of D_p follow from the general theorem plus the specific inequalities proved in Corollary 1. The paper does cite prior work, including [4], but not as a self-citation and not as the load-bearing justification for its own new claims. The one substantive concern is the parenthetical in Section 3.2 claiming that the Fubini–Study metric satisfies (31) for an arbitrary norm on B(H); for the Frobenius norm the inequality d_FS ≥ ‖·‖_Fr fails. This is a correctness/scope issue with the presentation of Theorem 3's hypotheses, not a circularity: Theorem 3 remains a valid conditional statement, and Corollary 1(ii) relies on the explicitly proved bound d_FS ≥ 2^{-1/2}‖·‖_Fr rather than on the false universal claim. No step in the derivation reduces to its own inputs by construction, so the circularity score is 0.
Assumptions & free parameters
assumptions (5)
- standard math Choquet–Bishop–De Leeuw theorem: every point of a compact convex subset of a locally convex Hausdorff space admits a representing probability on the extreme boundary.
- standard math Standard Wasserstein theory: for compact Polish (E,d), W_p metrizes weak-* convergence on P(E), and (P(E),W_p) is compact Polish and geodesic when (E,d) is geodesic.
- standard math Quotient metric space facts: the quotient of a compact metric space by a closed equivalence relation is compact and the quotient distance metrizes the quotient topology; quotient of a length space is a length space.
- domain assumption Finite-dimensional spectral theorem and the characterization of representing measures of a density matrix by dim H - 1 linear eigenvalue constraints.
- standard math Winkler/Dubins extreme-point theorem for convex sets of probability measures defined by finitely many linear constraints: extreme points have finite support.
Cite this review
Pith. "Pith review of Folded optimal transport and its application to separable quantum optimal transport." pith.science (2026). https://pith.science/paper/6HIHITNA
@misc{pith2026251201722,
author = {Pith},
title = {Pith review of: Folded optimal transport and its application to separable quantum optimal transport},
year = {2026},
howpublished = {\url{https://pith.science/paper/6HIHITNA}},
note = {Machine review of arXiv:2512.01722}
}
read the original abstract
We introduce folded optimal transport, as a method to extend a cost or distance defined on the extreme boundary of a convex to the whole convex, related to convex extension. This construction broadens the framework of standard optimal transport, found to be the particular case of the convex being a simplex. Relying on Choquet's theory and standard optimal transport, we introduce the folded Kantorovich cost and folded Wasserstein distances, and study their induced metric properties. We then apply the construction to the quantum setting, and obtain an actual separable quantum Wasserstein distance on the set of density matrices from a distance on the set of pure states, closely related to the semi-distance of Beatty and Stilck-Franca [4], and of which we obtain a variety of properties. We also find that the semiclassical Golse-Paul [16] cost writes as a folded Kantorovich cost. Folded optimal transport therefore provides a unified framework for classical, semiclassical and separable quantum optimal transport.
Reference graph
Works this paper leans on
-
[4]
Beatty and D
E. Beatty and D. Stilck França. Order p quantum Wasserstein distances from couplings. InAnn. Henri Poincaré, pages 1–59. Springer, 2025
2025
-
[1]
Alfsen.Compact convex sets and boundary integrals, volume 57
E.M. Alfsen.Compact convex sets and boundary integrals, volume 57. Springer Science & Business Media, 2012
2012
-
[2]
H. Bauer. Silovscher rand und dirichletsches problem. InAnn. Inst. Fourier, volume 11, pages 89–136, 1961
1961
-
[3]
E. Beatty. Wasserstein distances on quantum structures: an overview.arXiv preprint arXiv:2506.09794, 2025
arXiv 2025
-
[5]
A free probability analogue of the wasserstein metric on the trace-state space.Geom
Philippe Biane and Dan V oiculescu. A free probability analogue of the wasserstein metric on the trace-state space.Geom. Funct. Anal. GAFA, 11(6):1125–1138, 2001
2001
-
[6]
O. Bucicovschi and J. Lebl. On the continuity and regularity of convex extensions.arXiv preprint arXiv:1012.5796, 2010
arXiv 2010
-
[7]
Burago, Y
D. Burago, Y . Burago, and S. Ivanov.A course in metric geometry, volume 33. American Mathematical Society Providence, 2001
2001
-
[8]
Carlen and J
E.A. Carlen and J. Maas. An analog of the 2-Wasserstein metric in non-commutative probability under which the fermionic Fokker–Planck equation is gradient flow for the entropy.Commun. Math. Phys., 331(3):887–926, 2014
2014
Show all 30 references
-
[9]
Chitambar and G
E. Chitambar and G. Gour. Quantum resource theories.Rev. Mod. Phys., 91:025001, Apr 2019
2019
-
[10]
Connes and J
A. Connes and J. Lott. The metric aspect of noncommutative geometry. InNew symmetry principles in quantum field theory, pages 53–93. Springer, 1992
1992
-
[11]
Dantzig and M.N
G.B. Dantzig and M.N. Thapa.Linear programming: 2: theory and extensions. Springer, 2003
2003
-
[12]
De Palma and D
G. De Palma and D. Trevisan. Quantum optimal transport with quantum channels. InAnn. Henri Poincaré, volume 22, pages 3199–3234. Springer, 2021
2021
-
[13]
Delon and A
J. Delon and A. Desolneux. A Wasserstein-type distance in the space of Gaussian mixture models.SIAM J. Imaging Sci., 13(2):936–970, 2020
2020
-
[14]
L.E. Dubins. On extreme points of convex sets.Journal of Mathematical Analysis and Applications, 5(2):237–244, 1962
1962
-
[15]
Golse, C
F. Golse, C. Mouhot, and T. Paul. On the mean field and classical limits of quantum mechanics.Commun. Math. Phys., 343(1):165–205, 2016
2016
-
[16]
Golse and T
F. Golse and T. Paul. The Schrödinger equation in the mean-field and semiclassical regime.Arch. Ration. Mech. Anal., 223:57–94, 2017
2017
-
[17]
Golse and T
F. Golse and T. Paul. Semiclassical evolution with low regularity.J. Math. Pures Appl., 151:257–311, 2021
2021
-
[18]
Golse and T
F. Golse and T. Paul. Quantitative observability for the Schrödinger and Heisenberg equations: An optimal transport approach. Math. Models Methods Appl. Sci., 32(05):941–963, 2022
2022
-
[19]
Hiriart-Urruty and C
J.B. Hiriart-Urruty and C. Lemaréchal.Fundamentals of convex analysis. Springer Science & Business Media, 2004
2004
-
[20]
C.J. Peifer. Choquet theory. 2020
2020
-
[21]
Phelps.Lectures on Choquet’s theorem
R.R. Phelps.Lectures on Choquet’s theorem. Springer, 2002
2002
-
[22]
Savaré and G
G. Savaré and G. E. Sodini. A simple relaxation approach to duality for optimal transport problems in completely regular spaces. J. Convex Anal., 29(1):1–12, 2022
2022
-
[23]
Tóth and J
G. Tóth and J. Pitrik. Quantum Wasserstein distance based on an optimization over separable states.Quantum, 7:1143, 2023
2023
-
[24]
Tóth and J
G. Tóth and J. Pitrik. Quantum Wasserstein distance and its relation to several types of fidelities.arXiv preprint arXiv:2506.14523, 2025
2025 arXiv
-
[25]
Trevisan
D. Trevisan. Quantum optimal transport: an invitation.Boll. Unione Mat. Ital., 18(1):347–360, 2025
2025
-
[26]
A. Uhlmann. Roofs and convexity.Entropy, 12(7):1799–1832, 2010
2010
-
[27]
Villani.Optimal transport: old and new, volume 338
C. Villani.Optimal transport: old and new, volume 338. Springer, 2008
2008
-
[28]
G. Winkler. Extreme points of moment sets.Math. Oper. Res., 13(4):581–587, 1988
1988
-
[29]
M. Yan. Extension of convex function.arXiv preprint arXiv:1207.0944, 2012
2012 arXiv
-
[30]
X. Zhu, C. Zhang, Z. An, and B. Zeng. Unified framework for calculating convex roof resource measures.npj Quantum Inf., 11(1):56, 2025. APPENDIXA. SOME LEMMAS AND PROOFS In this appendix, I have recast a few lemmas and proofs. A.1.Lemmas in the general setting.We consider the ...
2025
Reviewed August 3, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.