OctoInformBeta

OpenAI preprint, September 2026 · §3. A metric and volume bound for connected supports

Lemma 3.1: Compact realized balls, compatible realization topologies, and finite stabilizers for free vertex actions

Finiteness for transportation balls on connected supports.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

A simplicial chain uses finitely many vertex supports. The geometric question is which connected supports can meet a prescribed transportation ball. Let Γ\Gamma be a connected, locally finite graph on a vertex set YY, and let nn be a positive integer. The graph metric dΓd_\Gamma assigns to two vertices the length of a shortest edge path between them.✦

Write V=ℝ(Y)V=\mathbb R^{(Y)} for the vector space of finitely supported real functions on YY. The coefficient sum Σ\Sigma and the Euclidean norm on VV are given by Σ(v)=∑y∈Yvy,‖v‖2=(∑y∈Yvy2)1/2,supp⁡(v)={y∈Y:vy≠0}.\Sigma(v)=\sum_{y\in Y}v_y, \qquad \lVert v\rVert_2=\left(\sum_{y\in Y}v_y^2\right)^{1/2}, \qquad \operatorname{supp}(v)=\{y\in Y:v_y\ne0\}.(1) A finitely supported probability measure p∈Vp\in V on YY is a vector with nonnegative coefficients and coefficient sum one, using Σ\Sigma from (1); denote the set of these measures by 𝒫f(Y)\mathcal P_f(Y).

For p,q∈𝒫f(Y)p,q\in\mathcal P_f(Y) and a function f:Y→ℝf:Y\to\mathbb R, the expectation of ff and the expectation difference are 𝔼p[f]=∑y∈Ypyf(y),Δp,q(f)=𝔼p[f]−𝔼q[f].\mathbb E_p[f]=\sum_{y\in Y}p_yf(y), \qquad \Delta_{p,q}(f)=\mathbb E_p[f]-\mathbb E_q[f].(2) The dual transportation distance W1W_1 on 𝒫f(Y)\mathcal P_f(Y) is the supremum of the differences in (2) over 11-Lipschitz functions for dΓd_\Gamma; its normalization ρ\rho is the metric used on 𝒫f(Y)\mathcal P_f(Y): W1(p,q)=supf:Y→ℝ|f(y)−f(z)|≤dΓ(y,z)for all y,zΔp,q(f),ρ(p,q)=W1(p,q)n.W_1(p,q)=\sup_{\substack{f:Y\to\mathbb R\\ |f(y)-f(z)|\le d_\Gamma(y,z)\ \text{for all }y,z}} \Delta_{p,q}(f), \qquad \rho(p,q)=\frac{W_1(p,q)}{\sqrt n}.(3) The closed transportation ball with center c∈𝒫f(Y)c\in\mathcal P_f(Y) and radius s≥0s\ge0 is Ls(c)={p∈𝒫f(Y):ρ(c,p)≤s}.L_s(c)=\{p\in\mathcal P_f(Y):\rho(c,p)\le s\}.(4)

Realizations and Vertex Actions

Let 𝒯\mathcal T be a family of finite subsets σ⊂Y\sigma\subset Y, each with n+1n+1 vertices and connected induced graph Γ[σ]\Gamma[\sigma]. The probability simplex PσP_\sigma, its affine coordinate simplex Δσ\Delta_\sigma, and the two realizations of 𝒯\mathcal T are Pσ={p∈𝒫f(Y):supp⁡(p)⊆σ},Δσ={v∈V:vy≥0for all y,Σ(v)=1,supp⁡(v)⊆σ},R𝒯=⋃σ∈𝒯Pσ,A𝒯=⋃σ∈𝒯Δσ.\begin{aligned} P_\sigma&=\{p\in\mathcal P_f(Y):\operatorname{supp}(p)\subseteq\sigma\},\\ \Delta_\sigma&=\{v\in V:v_y\ge0\ \text{for all }y,\ \Sigma(v)=1,\ \operatorname{supp}(v)\subseteq\sigma\},\\ R_{\mathcal T}&=\bigcup_{\sigma\in\mathcal T}P_\sigma, &A_{\mathcal T}&=\bigcup_{\sigma\in\mathcal T}\Delta_\sigma. \end{aligned}(5) In (5), R𝒯R_{\mathcal T} has the topology induced by ρ\rho, and A𝒯A_{\mathcal T} has the topology induced by the Euclidean norm. The affine probability a𝒯(v)a_{\mathcal T}(v) associated to v∈A𝒯v\in A_{\mathcal T} is the probability measure with coefficient vector vv.✦

The affine realization of an abstract simplicial complex KK on YY is denoted by |K|aff⊂V|K|_{\mathrm{aff}}\subset V. For a nonempty face σ\sigma of KK, its affine simplex κ\kappa is κ=conv⁡{ey:y∈σ},(ey)z={1z=y,0z≠y.\kappa=\operatorname{conv}\{e_y:y\in\sigma\}, \qquad (e_y)_z=\begin{cases}1&z=y,\\0&z\ne y.\end{cases}(6) The barycentric probability pK(v)p_K(v) of v∈|K|affv\in|K|_{\mathrm{aff}} has coefficient vector vv and support a face of KK. For a center c∈𝒫f(Y)c\in\mathcal P_f(Y), the distance-from-center function on VV is δc(v)={ρ(c,pK(v))v∈|K|aff,0v∉|K|aff.\delta_c(v)= \begin{cases} \rho(c,p_K(v))&v\in|K|_{\mathrm{aff}},\\ 0&v\notin|K|_{\mathrm{aff}}. \end{cases}(7)✦ The metric on |K|aff|K|_{\mathrm{aff}} is obtained from its barycentric probabilities; Ls(c)L_s(c) also denotes a closed ball in this realization when cc belongs to it.

Let a group HH act freely on YY, and let CC be a finite set indexing centers ci∈𝒫f(Y)c_i\in\mathcal P_f(Y). For a bijection e:Y→Ze:Y\to Z, the pushforward e*pe_*p is the probability measure on ZZ whose coefficient at zz is pe−1(z)p_{e^{-1}(z)}.✦ The induced action on probabilities and the setwise stabilizer of a finite subset F⊂YF\subset Y are (h⋅p)y=ph−1⋅y,HF={h∈H:hF=F}.(h\cdot p)_y=p_{h^{-1}\cdot y}, \qquad H_F=\{h\in H:hF=F\}.(8)

The Claim

Theorem 1. Let Γ\Gamma, nn, and 𝒯\mathcal T satisfy the conditions above. For every c∈𝒫f(Y)c\in\mathcal P_f(Y) and s≥0s\ge0, with Ls(c)L_s(c) as in (4), {σ∈𝒯:Pσ∩Ls(c)≠⌀}is finite.\{\sigma\in\mathcal T:P_\sigma\cap L_s(c)\ne\varnothing\} \quad\text{is finite}.(9)✦ Every p∈𝒫f(Y)p\in\mathcal P_f(Y) has a ρ\rho-open neighborhood OO such that {σ∈𝒯:O∩Pσ≠⌀}is finite.\{\sigma\in\mathcal T:O\cap P_\sigma\ne\varnothing\} \quad\text{is finite}.(10)✦ For every center cc and radius ss as above, R𝒯∩Ls(c)is compact.R_{\mathcal T}\cap L_s(c)\quad\text{is compact}.(11)✦ The coefficient map is a homeomorphism R𝒯→≅A𝒯,p↦(py)y∈Y,v↦a𝒯(v)for its inverse.R_{\mathcal T}\xrightarrow{\ \cong\ }A_{\mathcal T}, \qquad p\longmapsto(p_y)_{y\in Y}, \qquad v\longmapsto a_{\mathcal T}(v)\ \text{for its inverse}.(12)✦ For a free action of HH on YY as in (8), a finite family (ci)i∈C(c_i)_{i\in C}, any s≥0s\ge0, and any affine simplex κ\kappa of KK as in (6), {(h,i)∈H×C:{v∈V:δh⋅ci(v)≤s}∩κ≠⌀}is finite.\left\{(h,i)\in H\times C: \{v\in V:\delta_{h\cdot c_i}(v)\le s\}\cap\kappa\ne\varnothing\right\} \quad\text{is finite}.(13)✦ For a free action of HH on any set YY and every nonempty finite subset F⊂YF\subset Y, HFis finite.H_F\quad\text{is finite}.(14)✦

In the paper, with |K||K| denoting the realization identified by barycentric coordinates, the result reads: in the standing setting, every closed ρ\rho-ball meets only finitely many simplices of KK, the complex KK is locally finite, ρ\rho induces its usual realization topology and is proper, for c∈|K|c\in|K|, s≥0s\ge0, and every simplex κ\kappa the set {h∈H:Ls(hc)∩κ≠⌀}\{h\in H:L_s(hc)\cap\kappa\ne\varnothing\} is finite, and the setwise stabilizer HκH_\kappa of every nonempty simplex κ\kappa is finite.

What Is Checked

The assertions (9), (10), (11), and (12) are stated over finite probabilities on the graph metric and the connected supports in 𝒯\mathcal T; (13) uses the distance function (7) and finitely many centers under a free vertex action, while (14) is stated for an arbitrary free action.

Distinct vertices have graph distance at least one.✦ Let ℓ2(Y)\ell^2(Y) be the space of square-summable real functions on YY; the inclusion V→ℓ2(Y)V\to\ell^2(Y) is injective and induces the norm in (1).