Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 6.8 — δ_4k < √2 − 1 implies the strong rank null space property of order k with η = √2 δ_4k/(1 − δ_3k)

Proved
IRLSM.RIP.proposition_6_8

by mikedeng1 · Oct 7, 2026 · Mathlib 0df444a (Lean v4.33.1)

low-rank-recoverynull-space-propertyp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1restricted-isometry-property

Let S:Rn×p→Rm\mathcal S : \mathbb R^{n\times p}\to\mathbb R^mS:Rn×p→Rm be a linear map whose rank restricted isometry constants (Definition 1.1) satisfy δ4k>0\delta_{4k} > 0δ4k​>0 and

δ4k<2−1≈0.41.\delta_{4k} < \sqrt 2 - 1 \approx 0.41 .δ4k​<2​−1≈0.41.

Then S\mathcal SS satisfies the strong rank null space property of order kkk (Definition 6.4) with constant

η=2 δ4k1−δ3k∈(0,1).\eta = \sqrt 2\,\frac{\delta_{4k}}{1 - \delta_{3k}} \in (0,1).η=2​1−δ3k​δ4k​​∈(0,1).

That is, for every nonzero XXX in the kernel of S\mathcal SS and every decomposition X=X1+X2X = X_1 + X_2X=X1​+X2​ with rank⁡X1≤k\operatorname{rank}X_1 \le krankX1​≤k, there are H1,H2H_1, H_2H1​,H2​ with X=H1+H2X = H_1 + H_2X=H1​+H2​, rank⁡H1≤2k\operatorname{rank}H_1 \le 2krankH1​≤2k, ⟨H1,H2⟩=0\langle H_1, H_2\rangle = 0⟨H1​,H2​⟩=0, X1H2⊤=0X_1H_2^{\top} = 0X1​H2⊤​=0, X1⊤H2=0X_1^{\top}H_2 = 0X1⊤​H2​=0 and ∥H1∥∗≤η∥H2∥∗\|H_1\|_* \le \eta\|H_2\|_*∥H1​∥∗​≤η∥H2​∥∗​.

Together with the convergence analysis of the IRLS-M algorithm, this shows that the algorithm recovers every matrix of rank at most kkk from measurements with a small restricted isometry constant, and that nuclear-norm minimization does so stably.

Formalization Note. Real matrices; S(X)ℓ=⟨Aℓ,X⟩\mathcal S(X)_\ell = \langle A_\ell, X\rangleS(X)ℓ​=⟨Aℓ​,X⟩. δk\delta_kδk​ is the squared-form constant of Definition 1.1, defined for every kkk (here 4k4k4k may exceed nnn). The hypothesis δ4k>0\delta_{4k} > 0δ4k​>0 is Definition 1.1's own requirement δk>0\delta_k > 0δk​>0; without it η\etaη could be 0∉(0,1)0 \notin (0,1)0∈/(0,1). The conclusion includes η∈(0,1)\eta \in (0,1)η∈(0,1), which is part of Definition 6.4.

Preamble
import Mathlib
import Definitions.Def_IRLSM_Convergence_Basic

open HighDimStat.MatrixRank Matrix
Formal statement
namespace IRLSM.RIP

/-- Proposition 6.8: if the rank restricted isometry constant satisfies `δ_{4k} < √2 − 1`, then
`S` has the strong rank null space property of order `k` with constant
`η = √2 δ_{4k}/(1 − δ_{3k}) ∈ (0, 1)`.

Fornasier–Rauhut–Ward, arXiv:1010.2471v4, Proposition 6.8 with (6.6), p. 18.

Formalization Notes: real matrices; `S(X)_l = ⟨A_l, X⟩`; `δ_k` is `ripConst A k` (squared form
of Definition 1.1, defined for every `k`, since `4k` may exceed `n`). The hypothesis
`0 < δ_{4k}` is the positivity "δ_k > 0" of Definition 1.1; without it `η` could be
`0 ∉ (0, 1)`. `SRNSP` contains `0 < η < 1`, so the conclusion includes "η ∈ (0, 1)". -/
theorem proposition_6_8 {n p m : ℕ} (A : Fin m → Matrix (Fin n) (Fin p) ℝ) (k : ℕ)
    (hpos : 0 < IRLSM.Convergence.ripConst A (4 * k)) (h66 : IRLSM.Convergence.ripConst A (4 * k) < Real.sqrt 2 - 1) :
    IRLSM.Convergence.SRNSP A k (Real.sqrt 2 * IRLSM.Convergence.ripConst A (4 * k) / (1 - IRLSM.Convergence.ripConst A (3 * k))) := by sorry

end IRLSM.RIP
Source
Fornasier, Rauhut, Ward, Low-rank matrix recovery via iteratively reweighted least squares minimization, arXiv:1010.2471v4 (2011), Proposition 6.8 with (6.6), p. 18; proof pp. 19–20
Human review
  • Endorsed by Shuze Chen · Oct 7, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 7, 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