Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 6.12, p. 345 — for ξ ∼ N(0, Σ) with unit diagonal and ζ = sign(ξ), E ζ_iζ_j = (2/π) arcsin(Σ_ij)

Proved
ConvexOptAlg.GoemansWilliamson.lemma_6_12

by mikedeng1 · Oct 5, 2026 · Mathlib 0df444a (Lean v4.33.1)

gaussianp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1probabilityrandomized-roundingsheppard-formula

Let Σ∈Rn×n\Sigma\in\mathbb R^{n\times n}Σ∈Rn×n be symmetric positive semidefinite with Σi,i=1\Sigma_{i,i}=1Σi,i​=1 for every i∈[n]i\in[n]i∈[n]. Let ξ∼N(0,Σ)\xi\sim\mathcal N(0,\Sigma)ξ∼N(0,Σ) be a centered Gaussian vector in Rn\mathbb R^nRn with covariance matrix Σ\SigmaΣ, and let ζ=sign⁡(ξ)∈{−1,1}n\zeta=\operatorname{sign}(\xi)\in\{-1,1\}^nζ=sign(ξ)∈{−1,1}n. Then for all i,j∈[n]i,j\in[n]i,j∈[n], ζiζj\zeta_i\zeta_jζi​ζj​ is integrable and

E ζiζj=2πarcsin⁡(Σi,j).\mathbb E\,\zeta_i\zeta_j=\frac{2}{\pi}\arcsin(\Sigma_{i,j}).Eζi​ζj​=π2​arcsin(Σi,j​).

This is Sheppard's formula for the correlation of the signs of two jointly Gaussian variables with unit variances; it computes the expected value of the Goemans–Williamson rounding term by term.

Formalization Note N(0,Σ)\mathcal N(0,\Sigma)N(0,Σ) is Mathlib's multivariateGaussian 0 Σ on EuclideanSpace ℝ (Fin n), the image of the standard Gaussian under Σ1/2\Sigma^{1/2}Σ1/2; it is defined for singular Σ\SigmaΣ as well. Positive semidefiniteness of Σ\SigmaΣ is implicit in the book ("ξ∼N(0,Σ)\xi\sim\mathcal N(0,\Sigma)ξ∼N(0,Σ)": Σ\SigmaΣ is a covariance matrix) and is stated as a hypothesis, since Mathlib's construction returns the Dirac mass at 000 for a non-PSD matrix. The sign takes the value 111 at 000. Integrability of the bounded integrand is part of the conclusion.

Preamble
import Mathlib
import Definitions.Def_ConvexOptAlg_GoemansWilliamson_Defs
Formal statement
namespace ConvexOptAlg.GoemansWilliamson

open MeasureTheory ProbabilityTheory

/-- **Lemma 6.12** (Bubeck, arXiv:1405.4980v2, §6.6, p. 345). Let `ξ ∼ N(0, Σ)` with `Σ i i = 1`
for `i ∈ [n]`, and `ζ = sign(ξ)`. Then `E ζᵢζⱼ = (2/π) arcsin(Σ i j)`.

`N(0, Σ)` is Mathlib's `multivariateGaussian 0 Σ` on `EuclideanSpace ℝ (Fin n)` (the centered
Gaussian with covariance matrix `Σ`, defined for every positive semidefinite `Σ`, singular or
not); `Σ` positive semidefinite is implicit in the book (it is a covariance matrix). The sign
takes values in `{−1, 1}` (`sgn 0 = 1`). The integrand is bounded and measurable; its
integrability is asserted as part of the conclusion. -/
theorem lemma_6_12 {n : ℕ} (Sig : Matrix (Fin n) (Fin n) ℝ) (hpsd : Sig.PosSemidef)
    (hdiag : ∀ i, Sig i i = 1) (i j : Fin n) :
    Integrable (fun ξ : EuclideanSpace ℝ (Fin n) => sgn (ξ i) * sgn (ξ j))
        (multivariateGaussian 0 Sig) ∧
      ∫ ξ, sgn (ξ i) * sgn (ξ j) ∂(multivariateGaussian 0 Sig) =
        2 / Real.pi * Real.arcsin (Sig i j) := by sorry

end ConvexOptAlg.GoemansWilliamson
Source
Bubeck, arXiv:1405.4980v2, Lemma 6.12, p. 345
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me