OctoInformBeta

OpenAI preprint, September 2026 · §3. Finite affine potentials and prescribed cell masses

§3 (setting): Laguerre cells, masses, moments and face coefficients: Closed Laguerre cells and intrinsic measures for finite affine potentials

Cell masses and canonical face coefficients.

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

Let K⊂ℝdK\subset\mathbb R^d be a compact convex source body of positive Lebesgue volume, and let II be a finite nonempty index set. For sites yi∈ℝdy_i\in\mathbb R^d and intercepts ci∈ℝc_i\in\mathbb R, the finite affine potential uu is the function in (1): u(x)=maxi∈I(x⋅yi+ci).u(x)=\max_{i\in I}\bigl(x\cdot y_i+c_i\bigr).(1) The cell masses describe the distribution of source mass among the affine terms; the face coefficients use surface area and first moments along their interfaces.

Write ℒd\mathcal L^d for Lebesgue measure and |K|=ℒd(K)|K|=\mathcal L^d(K) for the volume of KK. The uniform measure ρ\rho on KK is the normalized restriction of Lebesgue measure given by (2): ρ=1|K|ℒd↾K.\rho=\frac{1}{|K|}\,\mathcal L^d\!\restriction_K.(2)✦

Intrinsic Measures

For a linear subspace W⊂ℝdW\subset\mathbb R^d and a point q∈ℝdq\in\mathbb R^d, the intrinsic affine plane measure ΠW,q\Pi_{W,q} is the pushforward of the Euclidean Lebesgue measure ℒW\mathcal L_W on WW under translation by qq, as in (3): ΠW,q=(w↦q+w)#ℒW.\Pi_{W,q}=(w\mapsto q+w)_\#\mathcal L_W.(3)✦

For a normal vector n∈ℝdn\in\mathbb R^d and an offset a∈ℝa\in\mathbb R, the canonical basepoint q(n,a)q(n,a) is the vector defined by (4): q(n,a)={−a∥n∥2n,n≠0,0,n=0.q(n,a)= \begin{cases} -\dfrac{a}{\|n\|^2}\,n,&n\ne0,\\ 0,&n=0. \end{cases}(4)✦ The intrinsic hyperplane measure λn,a\lambda_{n,a} is the translated subspace measure in (5), where n⟂n^\perp is the orthogonal complement of the span of nn: λn,a=Πn⟂,q(n,a).\lambda_{n,a}=\Pi_{n^\perp,q(n,a)}.(5)✦

The Definition

Definition 1. For the source KK and affine data (yi,ci)i∈I(y_i,c_i)_{i\in I}, the closed Laguerre cell PiP_i is the subset of KK defined by (6): Pi={x∈K:x⋅yj+cj≤x⋅yi+cifor every j∈I}.P_i=\left\{x\in K: x\cdot y_j+c_j\le x\cdot y_i+c_i\quad\text{for every }j\in I\right\}.(6)✦ The normalized cell mass MiM_i is the real number in (7): Mi=ρ(Pi).M_i=\rho(P_i).(7)✦

For i,j∈Ii,j\in I, the intrinsic interface measure σij\sigma_{ij} is the measure on ℝd\mathbb R^d defined by (8): σij=λyi−yj,ci−cj.\sigma_{ij}=\lambda_{y_i-y_j,\,c_i-c_j}.(8)✦ The canonical face weight gijg_{ij} is the normalized face area in (9): gij=σij(Pi∩Pj)|K|∥yi−yj∥.g_{ij}=\frac{\sigma_{ij}(P_i\cap P_j)}{|K|\,\|y_i-y_j\|}.(9)✦ The normalized face first moment bij∈ℝdb_{ij}\in\mathbb R^d is the vector in (10): bij=1|K|∥yi−yj∥∫Pi∩Pjxdσij(x).b_{ij}=\frac{1}{|K|\,\|y_i-y_j\|} \int_{P_i\cap P_j}x\,d\sigma_{ij}(x).(10)✦ In (9) and (10), a zero denominator has reciprocal zero.

What Is Checked

The cell definitions are stated over Euclidean source sets with their uniform measure and affine site data; the intrinsic surface measures are defined from Euclidean volume on linear subspaces.