pith. sign in

arxiv: 2604.17924 · v2 · pith:VVO7F4KQnew · submitted 2026-04-20 · 🧮 math.MG · math.OC

Wasserstein barycenters on metric graphs

J\'er\^ome Bertrand (IMT) , Jianyu Ma (IMT) This is my paper

Pith reviewed 2026-05-10 03:46 UTC · model grok-4.3

classification 🧮 math.MG math.OC
keywords Wasserstein barycentermetric graphabsolute continuityHausdorff measureoptimal transportbarycenters
0
0 comments X

The pith

Wasserstein barycenters on metric graphs are absolutely continuous away from vertices under appropriate conditions.

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

This paper establishes conditions under which the Wasserstein barycenter of measures supported on a metric graph is absolutely continuous with respect to the Hausdorff measure away from the graph's vertices. This property implies that the barycenter measure can be expressed via a density function along the edges of the graph. A reader might care because it describes how averaged optimal transport behaves on one-dimensional network spaces, avoiding unexpected concentrations of mass on edge interiors. If these conditions are satisfied, the barycenter behaves regularly on each connected component between vertices.

Core claim

The authors provide conditions for a Wasserstein barycenter to be absolutely continuous with respect to its Hausdorff measure away from the vertices of the graph.

What carries the argument

The Wasserstein barycenter of probability measures on the metric graph, defined as the minimizer of the weighted sum of squared Wasserstein distances, carrying the absolute continuity property on edge interiors.

Load-bearing premise

The abstract does not specify the precise assumptions on the metric graph, the input measures, or the barycenter weights.

What would settle it

Constructing a metric graph and input measures that satisfy the paper's conditions but yield a barycenter with an atom strictly in the interior of an edge would disprove the claim.

Figures

Figures reproduced from arXiv: 2604.17924 by J\'er\^ome Bertrand (IMT), Jianyu Ma (IMT).

Figure 1
Figure 1. Figure 1: P = P3 i=1 1 3 δνi on the tripod be a probability measure on the Wasserstein space over the tripod, where each νi is an absolutely continuous measure supported on the outer half [ 1 2 , 1] of a distinct branch. The unique Wasserstein barycenter of P is the Dirac measure µP = δ0 at the central vertex. More in general, the absence of synthetic lower curvature bounds on metric (measure) graphs forces to tackl… view at source ↗
read the original abstract

In this paper we provide conditions for a Wasserstein barycenter to be absolutely continuous with respect to its Hausdorff measure away from the vertices of the graph. One assumption removed in v2, the result is now optimal.

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 / 2 minor

Summary. The paper provides conditions under which the Wasserstein barycenter of probability measures on a finite connected compact metric graph (with finitely many edges) is absolutely continuous with respect to 1-Hausdorff measure on the interiors of the edges, away from vertices. Input measures are assumed to be probabilities with finite second moments. The central argument proceeds by disintegrating optimal transport plans along edges and invoking convexity of the Wasserstein functional to obtain the absolute continuity result.

Significance. If the stated conditions hold, the result supplies a useful regularity statement for Wasserstein barycenters on graphs, extending known absolute-continuity properties from Euclidean or Riemannian settings to a discrete network geometry. The derivation via disintegration and convexity is a standard, direct approach that appears internally consistent for the regime considered and could support further work on optimal transport on length spaces.

minor comments (2)
  1. [Abstract and §1] The abstract states that conditions are provided but does not list them; the introduction and main theorem statements should explicitly restate the graph and measure assumptions (finite connected compact length space, finitely many edges, finite second moments) so that the precise regime is visible without reading the full proofs.
  2. [Main theorem proof (likely §3 or §4)] In the disintegration step along edges, the notation for the conditional measures and the reference to convexity of the Wasserstein functional could be expanded with a short inline reminder of the relevant convexity inequality to aid readers who are not specialists in optimal transport on graphs.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive evaluation of our manuscript and the recommendation of minor revision. The referee's summary correctly identifies the main result: conditions ensuring that Wasserstein barycenters of probability measures with finite second moments on a finite connected compact metric graph are absolutely continuous with respect to 1-Hausdorff measure on the interiors of the edges, away from vertices. The derivation via disintegration of optimal plans and convexity of the Wasserstein functional is indeed the approach taken.

Circularity Check

0 steps flagged

No significant circularity; derivation self-contained

full rationale

The central claim is established by disintegrating optimal transport plans along the edges of the finite metric graph and invoking convexity of the Wasserstein functional, under explicitly stated assumptions on the graph (finite, connected, compact length space) and measures (probability measures with finite second moments). No step reduces by definition to its own inputs, no fitted parameters are relabeled as predictions, and no load-bearing premise rests on self-citation chains or imported uniqueness theorems. The argument uses standard optimal-transport tools without internal circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; therefore no free parameters, axioms, or invented entities can be extracted from the full derivation.

pith-pipeline@v0.9.0 · 5305 in / 896 out tokens · 40319 ms · 2026-05-10T03:46:07.611677+00:00 · methodology

discussion (0)

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