Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Law of a block V1TΩV_1^{\mathsf T}\OmegaV1T​Ω of a Gaussian matrix under an orthonormal frame

Proved
GaussianMatrix.block_law

by tc · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

gaussian-matrixprobabilityrandom-matrix-theoryrandomized-numerical-linear-algebra

Let Ω∈Rn×t\Omega\in\mathbb R^{n\times t}Ω∈Rn×t be a standard Gaussian matrix and let V1∈Rn×kV_1\in\mathbb R^{n\times k}V1​∈Rn×k have orthonormal columns, V1TV1=IkV_1^{\mathsf T}V_1=I_kV1T​V1​=Ik​. Then

V1TΩ ∼ γk,t,V_1^{\mathsf T}\Omega\ \sim\ \gamma_{k,t},V1T​Ω ∼ γk,t​,

i.e. V1TΩV_1^{\mathsf T}\OmegaV1T​Ω is a standard Gaussian k×tk\times tk×t matrix.

In the analysis of the randomized SVD one writes the right singular vectors of the input as [V1 V2][V_1\ V_2][V1​ V2​] and works with Ω1=V1TΩ\Omega_1=V_1^{\mathsf T}\OmegaΩ1​=V1T​Ω; this statement says Ω1\Omega_1Ω1​ is itself standard Gaussian, so the moment identities and singular-value tail bounds for Gaussian matrices apply to it directly.

Formalization Note. The statement is an equality of measures: the pushforward of γn,t\gamma_{n,t}γn,t​ under Ω↦V1TΩ\Omega\mapsto V_1^{\mathsf T}\OmegaΩ↦V1T​Ω equals γk,t\gamma_{k,t}γk,t​. The hypothesis V1TV1=IkV_1^{\mathsf T}V_1=I_kV1T​V1​=Ik​ forces k≤nk\le nk≤n.

Preamble
import Definitions.Def_GaussianMatrix_basic

open MeasureTheory ProbabilityTheory
open scoped Matrix
Formal statement
namespace GaussianMatrix
theorem block_law {n k t : ℕ} (V₁ : Matrix (Fin n) (Fin k) ℝ) (hV₁ : V₁ᵀ * V₁ = 1) :
    Measure.map (fun G : Fin n → Fin t → ℝ => Matrix.of.symm (V₁ᵀ * Matrix.of G))
      (gaussianMatrix n t) = gaussianMatrix k t := by sorry
end GaussianMatrix
Source
N. Halko, P.-G. Martinsson, J. A. Tropp, *Finding structure with randomness: probabilistic algorithms for constructing approximate matrix decompositions*, SIAM Review 53(2), 2011, https://arxiv.org/abs/0909.4061 (v2), §10.2 p. 57, proof of Theorem 10.5: "The Gaussian distribution is rotationally invariant, so V*Ω is also a standard Gaussian matrix. Observe that Ω₁ and Ω₂ are nonoverlapping submatrices of V*Ω, so these two matrices are not only standard Gaussian but also stochastically independent." (first claim)
Read-back

What the Lean code literally says, in plain math · claude-fable-5-1

Let n,k,tn, k, tn,k,t be natural numbers (implicit; any may be 000) and let V1∈Rn×kV_1 \in \mathbb{R}^{n \times k}V1​∈Rn×k be a real matrix with

V1TV1=IkV_1^{\mathsf T} V_1 = I_kV1T​V1​=Ik​

(orthonormal columns). Let γn,t\gamma_{n,t}γn,t​ be the law of an n×tn \times tn×t array with independent N(0,1)\mathcal{N}(0,1)N(0,1) entries (product of ntntnt copies of N(0,1)\mathcal{N}(0,1)N(0,1)), and γk,t\gamma_{k,t}γk,t​ the analogous law for k×tk \times tk×t arrays. The statement asserts the equality of measures on the space of k×tk \times tk×t real arrays

(G↦V1TG)# γn,t  =  γk,t,\big(G \mapsto V_1^{\mathsf T} G\big)_{\#}\,\gamma_{n,t} \;=\; \gamma_{k,t},(G↦V1T​G)#​γn,t​=γk,t​,

i.e. the pushforward of γn,t\gamma_{n,t}γn,t​ under the map taking an n×tn \times tn×t array GGG (viewed as a matrix) to the k×tk \times tk×t matrix V1TGV_1^{\mathsf T} GV1T​G (viewed as an array) equals γk,t\gamma_{k,t}γk,t​. Mathlib's pushforward would be the zero measure for a non-a.e.-measurable map, but this map is linear, hence measurable, so that convention is not triggered. Edge cases and vacuity: the hypothesis V1TV1=IkV_1^{\mathsf T}V_1 = I_kV1T​V1​=Ik​ forces k≤nk \le nk≤n (the rank of V1TV1V_1^{\mathsf T}V_1V1T​V1​ is at most nnn), so for k>nk > nk>n no such V1V_1V1​ exists and the statement is vacuous. If k=0k = 0k=0 the hypothesis is the trivial identity of 0×00 \times 00×0 matrices, satisfied by the unique n×0n \times 0n×0 matrix, and γ0,t\gamma_{0,t}γ0,t​ is the point mass on the one-point array space. If t=0t = 0t=0 both sides are point masses on one-point spaces. If n=0n = 0n=0 then kkk must be 000 for the hypothesis to hold.

Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by tc · Oct 9, 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