Prove2Me
Navigate
DiscoverFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

centered_sampling_independent_copy_rademacher_symmetrization_of_sample_ratio

Proved

by Shuze Chen · Jun 21, 2026 · Mathlib 0df444a (Lean v4.33.1)

bernoulli-samplingcandes-rechtlean4matrix-completionrademachersection-6-1symmetrization

This is the sample-ratio-safe Rademacher symmetrization step for the independent-copy difference in Candes-Recht Section 6.1.

Let

p=mn1n2,SΩ(X)=p−1(PΩ−pI)X.p=\frac{m}{n_1n_2},\qquad S_\Omega(X)=p^{-1}(P_\Omega-pI)X.p=n1​n2​m​,SΩ​(X)=p−1(PΩ​−pI)X.

Assume 0<n10<n_10<n1​, 0<n20<n_20<n2​, m≤n1n2m\le n_1n_2m≤n1​n2​, and q≥1q\ge1q≥1. If Ω\OmegaΩ and Ω′\Omega'Ω′ are independent Bernoulli samples with inclusion probability ppp, then the symmetric difference δij−δij′\delta_{ij}-\delta'_{ij}δij​−δij′​ may be represented by an independent Rademacher sign. In the theorem interface this gives the moment comparison

EΩ,Ω′ ∥SΩ(X)−SΩ′(X)∥q≤EΩ,Ω′Eε ∥SΩ,ε(X)−SΩ′,ε(X)∥q,\mathbb E_{\Omega,\Omega'}\,\|S_\Omega(X)-S_{\Omega'}(X)\|^q \le \mathbb E_{\Omega,\Omega'}\mathbb E_\varepsilon\,\|S_{\Omega,\varepsilon}(X)-S_{\Omega',\varepsilon}(X)\|^q,EΩ,Ω′​∥SΩ​(X)−SΩ′​(X)∥q≤EΩ,Ω′​Eε​∥SΩ,ε​(X)−SΩ′,ε​(X)∥q,

where SΩ,ε(X)=p−1∑(i,j)∈ΩεijXijeiej⊤S_{\Omega,\varepsilon}(X)=p^{-1}\sum_{(i,j)\in\Omega}\varepsilon_{ij}X_{ij}e_ie_j^\topSΩ,ε​(X)=p−1∑(i,j)∈Ω​εij​Xij​ei​ej⊤​.

Source: Candes-Recht 2008, PDF p. 24, Section 6.1, immediately after equation (6.5), where the symmetry of δab−δab′\delta_{ab}-\delta'_{ab}δab​−δab′​ introduces the Rademacher sequence.

Preamble
import Definitions.Def_matrix_completion_rademacher
open MatrixCompletion
Formal statement
theorem centered_sampling_independent_copy_rademacher_symmetrization_of_sample_ratio :
    ∀ (n₁ n₂ m q : ℕ) (X : Matrix (Fin n₁) (Fin n₂) ℝ),
      0 < n₁ → 0 < n₂ → m ≤ n₁ * n₂ → 1 ≤ q →
      bernoulliPairExpectation ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ)))
          (fun Omega Omega' =>
            spectralNorm
              (centeredSamplingFluctuation Omega
                  ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ))) X -
                centeredSamplingFluctuation Omega'
                  ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ))) X) ^ q) ≤
        bernoulliPairExpectation ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ)))
          (fun Omega Omega' =>
            rademacherExpectation
              (fun eps =>
                spectralNorm
                  (rademacherSampledMatrix Omega eps
                      ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ))) X -
                    rademacherSampledMatrix Omega' eps
                      ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ))) X) ^ q)) := by
  sorry
Source
Candes, Emmanuel, and Benjamin Recht. "Exact matrix completion via convex optimization." Communications of the ACM 55.6 (2012): 111-119.

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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me