REVIEW 2 major objections 3 minor 8 references
On averaged self-distances in finite dimensional Banach spaces
T0 review · 2 major / 3 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read A universal, dimension-dependent upper bound for averaged self-distances in finite-dimensional Banach spaces.
desk verdict Genuine new bound with a repairable algebraic error in Theorem 3; deserves peer review after a fix. 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 machinery is a covering of the unit ball K by homothetic copies rK. If K is covered by s such copies, then dividing the measure into the pieces L_i = (K ∩ K_i) \ (K_1 ∪ ⋯ ∪ K_{i−1}) gives an elementary estimate Δ(ν) ≤ 2 − 2(1 − r)/s. Choosing r optimally and using the Rogers–Zong bound on the number of translates needed, s ≈ (1 + $r^{{−1}}$)^n Θ_n with Θ_n < n log n + n log log n + 5n, yields the explicit f(n).
What would settle it
Find a Borel probability measure on a two-dimensional unit ball whose averaged self-distance exceeds (48 + 6√2)/31; if one exists, Theorem 2 is false. Alternatively, for any n ≥ 3, numerically maximize Δ(ν) over atomic measures on the cube and check whether any value exceeds 2(1 − $2^{{−n}}$ f(n)) with the explicit f(n) from the proof.
Extended reading notes
Core claim
The central claim is that for every Borel probability measure on the unit ball K of any n-dimensional real Banach space, the averaged self-distance Δ(ν) = ∫∫ |x − y| dν(x) dν(y) is bounded above by 2(1 − $2^{{−n}}$ f(n)), with f(n) an explicit universal function. The proof works by covering K by homothetic copies rK, partitioning the measure over the pieces, and then bounding distances within each piece by the diameter of that piece. For n = 2 the covering comes from Lassak's theorem that a plane convex body is covered by four homothetic copies of ratio √2/2; for n ≥ 3 it comes from Rogers–Zong translative covering estimates. The author also conjectures that the optimal bound is 2(1 − $2^{{−n}}$), attained when K is a parallelepiped and ν is uniform on its vertices.
Load-bearing premise
The numerical form of the bound depends on external theorems that fix how many small homothetic copies of a convex body suffice to cover it; if Lassak's four-copy result or the Rogers–Zong covering count failed, the stated f(n) would change.
Editorial extensions
If this is right
- The bound holds for every norm and every Borel measure, so it applies to any random pair of points drawn from the unit ball in any finite-dimensional normed space.
- For every n ≥ 2, the averaged self-distance is strictly below 2, ruling out measures that concentrate almost all mass on pairs of opposite extreme points.
- The explicit f(n) gives a concrete, computable upper bound for each n without hidden constants, though the paper notes the bound is crude for small dimensions.
- If the conjectured sharp bound 2(1 − 2^{−n}) is true, the maximum would be attained exactly by the cube with uniform vertex measure, giving a continuous analogue of Hadwiger's conjecture.
Reading between the lines
- Beyond the paper, a natural test is whether the method can be pushed to f(n) = 1 by using a covering by exactly 2^n homothetic copies of ratio 1/2, which would require a covering result stronger than the Rogers–Zong density estimates.
- The dimensional decay f(n) ∼ 2/(e n² log n) suggests that the averaged self-distance approaches 2 very quickly as dimension grows, leaving only an exponentially small gap; this asymptotic regime could be explored numerically for specific spaces like ℓ_p^n.
- The connection to Hadwiger's conjecture points to a probabilistic reformulation: the maximum over measures of the averaged self-distance selects geometries with minimal covering complexity, potentially linking extremal measures to tiling bodies.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The paper studies the maximal possible value of the averaged self-distance Δ(ν)=∫∫|x−y| dν⊗ν for Borel probability measures on the unit ball K of an n-dimensional real Banach space. It proves a planar bound Δ≤48/31+6√2/31 using Lassak's four-copy covering theorem, and for n≥3 it claims the bound Δ<2(1−2^{-n}·2(1−2/n)/(e n Θ_n)), where Θ_n is a universal upper bound for translative covering densities, yielding an explicit f(n) with f(n)∼2/(e n^2 log n). The paper also conjectures that the optimal general bound is Δ≤2(1−2^{-n}), with equality for parallelepipeds endowed with the uniform measure on their vertices.
Significance. If fully established, the main result would give a clean dimension-dependent improvement over the trivial bound Δ≤2 and would match the extremal example of the cube with uniform vertex measure, providing a continuous analogue of Hadwiger's conjecture. The strategy is economical and relies on standard external results (Lassak, Rogers, Rogers–Zong) that are correctly cited. The planar proof is sound. The higher-dimensional proof contains a localized algebraic error in the handling of the covering number; this error is repairable and does not destroy the asymptotic form of the main bound, but the printed derivation of Theorem 3 is not correct as it stands.
major comments (2)
- [Theorem 3, proof, first paragraph] The covering number is misstated. The proof says that K can be covered by at most s=(1+r^{-1})^n copies of rK, but the Rogers–Zong lemma stated immediately before gives at most (1+r^{-1})^n Θ(H)≤(1+r^{-1})^n Θ_n copies. Because of this omission, the displayed chain in the proof is algebraically false: if s=(1+r^{-1})^n, the final term should not contain Θ_n, while if s is intended to include Θ_n, then 1/s equals (1+r^{-1})^{-n}/Θ_n rather than (1+r^{-1})^{-n}Θ_n. The derivation as printed therefore does not establish Theorem 3.
- [Theorem 3, proof, final displayed inequality] Even after correcting the placement of Θ_n, the final estimate ((n−2)/(n−1))^n < (1−2/n)/e is false for every n≥3; for example, when n=3 it asserts 1/8 > 1/(3e), which is false. The valid elementary bound is ((n−2)/(n−1))^n < e^{-1}, which leads to Δ(ν)<2(1−2^{1-n}/(n e Θ_n))=2(1−2^{-n}·2/(e n Θ_n)). This is weaker than the bound stated in Theorem 3, so the exact statement of Theorem 3 must be weakened or the proof must be supplemented with a sharper argument. The asymptotic behavior f(n)∼2/(e n^2 log n) survives the correction.
minor comments (3)
- [Theorem 3, proof, notation for the number of copies] The notation s=(1+r^{-1})^n followed by K_1,...,K_{\lfloor s\rfloor} is ambiguous when s is not an integer; it would be clearer to introduce an integer m for the actual number of covering sets and to state the available integer upper bound m≤(1+r^{-1})^nΘ_n before applying the Cauchy inequality.
- [Theorem 2, proof, self-referential step] The sentence explaining how the bound κ_2 is applied to the restricted measures on the scaled copies K_i is terse; explicitly writing that a restriction of ν to L_i has double integral at most (√2/2)κ_2 v_i^2 because K_i is a homothetic copy of K of ratio √2/2 would make the displayed inequality easier to verify.
- [Heading and title] In the supplied text the running title contains spacing errors ('ON A VERAGED SELF-DISTANCES', 'SP ACES'); these should be corrected in the final version.
Circularity Check
No circularity: the proof is self-contained given external covering theorems and does not reduce its conclusions to its inputs.
full rationale
The derivation chain in this paper is not circular. The main results, Theorems 2 and 3, are obtained by combining an elementary variance-type inequality for partition weights with external covering results: Lassak's theorem for n = 2 and the Rogers and Rogers-Zong covering estimates for n >= 3. None of these inputs is defined in terms of the target quantity Delta(nu), and none is fitted to the data being predicted. The function f(n) is a concrete, parameter-free expression derived from the Rogers-Zong covering density; it is not a fitted parameter renamed as a prediction. The self-referential inequality in Theorem 2, where kappa_2 is bounded in terms of kappa_2, is a fixed-point argument that yields an explicit constant without assuming the conclusion, so it is not circular. The paper contains no load-bearing self-citations: all cited covering theorems are prior independent results by other authors. The concluding conjecture that f(n) might be replaced by 1 is clearly labelled as a hope and is not used in the proof. There is a possible algebraic slip in the printed proof of Theorem 3 concerning the placement of the factor Theta_n, but that is a correctness concern, not a circularity, and it does not affect the circularity score.
Assumptions & free parameters
assumptions (3)
- standard math Lassak's covering theorem: every plane convex body can be covered by four homothetic copies of ratio √2/2.
- standard math Rogers' covering density bound: for n≥3, every centrally symmetric convex body H satisfies Θ(H) ≤ Θ_L(H) < n log n + n log log n + 5n.
- standard math Rogers-Zong covering lemma: if 0<r<1, H can be covered by at most (1+r^{-1})^n Θ(H) translates of rH.
Cite this review
Pith. "Pith review of On averaged self-distances in finite dimensional Banach spaces." pith.science (2026). https://pith.science/paper/VWX7AVUY
@misc{pith2026241114129,
author = {Pith},
title = {Pith review of: On averaged self-distances in finite dimensional Banach spaces},
year = {2026},
howpublished = {\url{https://pith.science/paper/VWX7AVUY}},
note = {Machine review of arXiv:2411.14129}
}
abstract
Assume that $\mathfrak A$ is a real Banach space of finite dimension $n\geq2$. Consider any Borel probability measure $\nu$ supported on the unit ball $K$ of $\mathfrak A$. We show that \[\Delta(\nu)=\int_{x \in K}\int_{ y\in K}|x-y|_{\mathfrak A} \,\,\,\nu(x)\,\nu(y)\leq 2(1-2^{-n}f(n)),\] where $f:\mathbb N\setminus \{0,1\}\rightarrow (0,1]$ is a concrete universal function such that $f(n)\sim \frac{2}{\mathrm e n^2\log n}$. It is hoped that in the estimate`$f(n)$' can be replaced by `$1$'.
Reference graph
Works this paper leans on
-
[1]
Mathematika, 45 (1998), 245–252
Arias-de-Reyna, Juan; Ball, Keith; Villa, Rafael: Conc entration of the distance in finite dimensional normed spaces. Mathematika, 45 (1998), 245–252
work page 1998
-
[2]
: Excursions into Combinatorial Geometry
Boltyanski, Vladimir; Martini, Horst; Soltan, Petru S. : Excursions into Combinatorial Geometry. Universitext. Springer-Verlag, Berlin, 1997
work page 1997
-
[3]
Fejes T´ oth, G´ abor: A note on covering by convex bodies. Can. Math. Bull. 52 (2009), 361–365
2009
-
[4]
Hadwiger H.: Ungel¨ oste Probleme: No. 20. Elem. Math. 12 (1957), 121
work page 1957
-
[5]
Lassak, Marek: Covering a plane convex body by four homot hetical copies with the smallest positive ratio. Geom. Dedicata 21 (1986), 157–167
work page 1986
-
[6]
In: New Trends in Intuitive Geometry
Nasz´ odi, M´ arton: Flavors of translative coverings. In: New Trends in Intuitive Geometry. Springer Verlag, New York, Berlin, Heidelberg, 2018. Pages 335–358
work page 2018
-
[7]
A.: A note on coverings
Rogers, C. A.: A note on coverings. Mathematika 4 (1957), 1–6
1957
-
[8]
A.; Zong C.: Covering convex bodies by translat es of convex bodies
Rogers C. A.; Zong C.: Covering convex bodies by translat es of convex bodies. Mathematika 44 (1997), 215–218. Alfr´ed R ´enyi Institute of Mathematics, Re ´altanoda utca 13-15, Budapest, H–1053, Hun- gary Email address : lakos@renyi.hu
work page 1997
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.