Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Corollary 3.1, p. 17 — ‖Θ̂ − Θ*‖∞ ≲ r₁ + r₄ (blocks (B.1)–(B.3))

Proved
LagrangeRank.Debias.corollary_3_1

by mikedeng1 · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

constrained-inferencedebiased-estimatormatrix-perturbationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let c1,c2,L2∈Rc_1,c_2,L_2\in\mathbb Rc1​,c2​,L2​∈R and c3>0c_3>0c3​>0. There are constants C,δ>0C,\delta>0C,δ>0, depending only on c1,c2,c3,L2c_1,c_2,c_3,L_2c1​,c2​,c3​,L2​ and not on the dimension, such that the following holds for every ddd.

Let Ω^,Ω∗∈Rd×d\widehat\Omega,\Omega^*\in\mathbb R^{d\times d}Ω,Ω∗∈Rd×d and g^,g∗∈Rd\widehat g,g^*\in\mathbb R^dg​,g∗∈Rd, and let ρ1,ρ4≥0\rho_1,\rho_4\ge0ρ1​,ρ4​≥0 with ρ1+ρ4≤δ\rho_1+\rho_4\le\deltaρ1​+ρ4​≤δ. Suppose

∥g∗∥1≤c1,∥g^−g∗∥1≤L2ρ1,∥Ω∗∥∞≤c2,g∗⊤Ω∗g∗≥c3,∥Ω^−Ω∗∥∞≤ρ4.\|g^*\|_1\le c_1,\quad \|\widehat g-g^*\|_1\le L_2\rho_1,\quad \|\Omega^*\|_\infty\le c_2,\quad g^{*\top}\Omega^*g^*\ge c_3,\quad \|\widehat\Omega-\Omega^*\|_\infty\le\rho_4 .∥g∗∥1​≤c1​,∥g​−g∗∥1​≤L2​ρ1​,∥Ω∗∥∞​≤c2​,g∗⊤Ω∗g∗≥c3​,∥Ω−Ω∗∥∞​≤ρ4​.

Then, with Θ^=Θ(Ω^,g^)\widehat\Theta=\Theta(\widehat\Omega,\widehat g)Θ=Θ(Ω,g​) and Θ∗=Θ(Ω∗,g∗)\Theta^*=\Theta(\Omega^*,g^*)Θ∗=Θ(Ω∗,g∗) the block-inverse matrices of p. 16,

∣Θ^22−Θ22∗∣≤C(ρ1+ρ4),∥Θ^12−Θ12∗∥∞≤C(ρ1+ρ4),∥Θ^11−Θ11∗∥∞≤C(ρ1+ρ4).|\widehat\Theta_{22}-\Theta^*_{22}|\le C(\rho_1+\rho_4),\qquad \|\widehat\Theta_{12}-\Theta^*_{12}\|_\infty\le C(\rho_1+\rho_4),\qquad \|\widehat\Theta_{11}-\Theta^*_{11}\|_\infty\le C(\rho_1+\rho_4).∣Θ22​−Θ22∗​∣≤C(ρ1​+ρ4​),∥Θ12​−Θ12∗​∥∞​≤C(ρ1​+ρ4​),∥Θ11​−Θ11∗​∥∞​≤C(ρ1​+ρ4​).

In the paper, g^=∇f(θ^)\widehat g=\nabla f(\widehat\theta)g​=∇f(θ), g∗=∇f(θ∗)g^*=\nabla f(\theta^*)g∗=∇f(θ∗), ρ1≍r1\rho_1\asymp r_1ρ1​≍r1​ and ρ4≍r4\rho_4\asymp r_4ρ4​≍r4​, so this is Corollary 3.1, ∥Θ^−Θ∗∥∞≲r1+r4\|\widehat\Theta-\Theta^*\|_\infty\lesssim r_1+r_4∥Θ−Θ∗∥∞​≲r1​+r4​, on the three blocks (B.1)–(B.3) that its proof bounds. These are the blocks the proofs of Lemma A.1 and Theorem 3.1 use.

Formalization Note. The ≲\lesssim≲ is made quantitative: CCC and δ\deltaδ are chosen after c1,c2,c3,L2c_1,c_2,c_3,L_2c1​,c2​,c3​,L2​ and before ddd. The smallness ρ1+ρ4≤δ\rho_1+\rho_4\le\deltaρ1​+ρ4​≤δ is the paper's "recall r1,r4=o(1)r_1,r_4=o(1)r1​,r4​=o(1)" (p. 43). The hypothesis ∥g^−g∗∥1≤L2ρ1\|\widehat g-g^*\|_1\le L_2\rho_1∥g​−g∗∥1​≤L2​ρ1​ is Assumption 3.3 at t=1t=1t=1 combined with Assumption 3.1. The printed claim is the ∞\infty∞-norm of the whole (d+1)×(d+1)(d+1)\times(d+1)(d+1)×(d+1) difference; its last row contains ∥Θ^12−Θ12∗∥1\|\widehat\Theta_{12}-\Theta^*_{12}\|_1∥Θ12​−Θ12∗​∥1​, which needs column sums of Ω^\widehat\OmegaΩ that Assumption 3.4 does not control, and the proof (B.1)–(B.3) bounds only the three blocks stated here.

Preamble
import Mathlib
import Definitions.Def_LagrangeRank_Debias_Setting
Formal statement
namespace LagrangeRank.Debias
open Matrix
theorem corollary_3_1 (c1 c2 c3 L2 : ℝ) (hc3 : 0 < c3) :
    ∃ C δ : ℝ, 0 < C ∧ 0 < δ ∧
      ∀ (d : ℕ) (OmHat OmStar : Matrix (Fin d) (Fin d) ℝ) (gHat gStar : Fin d → ℝ)
        (ρ1 ρ4 : ℝ),
        0 ≤ ρ1 → 0 ≤ ρ4 → ρ1 + ρ4 ≤ δ →
        l1Norm gStar ≤ c1 → l1Norm (gHat - gStar) ≤ L2 * ρ1 →
        rowSumNorm OmStar ≤ c2 → c3 ≤ gStar ⬝ᵥ (OmStar *ᵥ gStar) →
        rowSumNorm (OmHat - OmStar) ≤ ρ4 →
        |(thetaBlock OmHat gHat - thetaBlock OmStar gStar).toBlocks₂₂ () ()|
            ≤ C * (ρ1 + ρ4) ∧
        ‖fun k => (thetaBlock OmHat gHat - thetaBlock OmStar gStar).toBlocks₁₂ k ()‖
            ≤ C * (ρ1 + ρ4) ∧
        rowSumNorm (thetaBlock OmHat gHat - thetaBlock OmStar gStar).toBlocks₁₁
            ≤ C * (ρ1 + ρ4) := by sorry
end LagrangeRank.Debias
Source
Liu, Fang, Lu, Lagrangian Inference for Ranking Problems, arXiv:2110.00151v1, p. 17, Corollary 3.1; proof in Appendix B.1, pp. 43–44, (B.1)–(B.3)
Human review
  • Endorsed by Shuze Chen · Oct 23, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 23, 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