Prove2Me
Navigate
DiscoverFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Theorem 2.4, proof, p. 4 — strong conic duality: 1=min⁡{⟨w0,z⟩:z−π∗(c)∈K∗, z∈L0⊥}1 = \min\{\langle w_0,z\rangle : z-\pi^*(c)\in K^*,\ z\in L_0^\perp\}1=min{⟨w0​,z⟩:z−π∗(c)∈K∗, z∈L0⊥​}, attained

Proved
ConeLifts.Factorization.slater_certificate

by mikedeng1 · Sep 27, 2026 · Mathlib 0df444a (Lean v4.33.1)

cone-liftsconic-dualityp2o-batch-p200ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Let n≥1n \ge 1n≥1, let C⊆RnC \subseteq \mathbb R^nC⊆Rn be a convex body and K⊆RmK \subseteq \mathbb R^mK⊆Rm a full-dimensional closed convex cone. Suppose C=π(K∩L)C = \pi(K \cap L)C=π(K∩L) for a linear map π:Rm→Rn\pi : \mathbb R^m \to \mathbb R^nπ:Rm→Rn and an affine subspace L=w0+L0L = w_0 + L_0L=w0​+L0​ (L0L_0L0​ its direction) with w0∈L∩int⁡Kw_0 \in L \cap \operatorname{int} Kw0​∈L∩intK. Let ccc be an extreme point of C∘C^\circC∘ and π∗\pi^*π∗ the adjoint of π\piπ. Then

1=min⁡{ ⟨w0,z⟩  :  z−π∗(c)∈K∗, z∈L0⊥ },1 = \min\{\, \langle w_0, z\rangle \;:\; z - \pi^*(c) \in K^*,\ z \in L_0^{\perp} \,\},1=min{⟨w0​,z⟩:z−π∗(c)∈K∗, z∈L0⊥​},

and the minimum is attained: every feasible zzz has ⟨w0,z⟩≥1\langle w_0, z\rangle \ge 1⟨w0​,z⟩≥1, and some feasible zzz has ⟨w0,z⟩=1\langle w_0, z\rangle = 1⟨w0​,z⟩=1.

This is the dual certificate from which the proof of Theorem 2.4 builds the factor B(c)=z−π∗(c)∈K∗B(c) = z - \pi^*(c) \in K^*B(c)=z−π∗(c)∈K∗. The underlying primal problem is max⁡{⟨π∗(c),w⟩:w∈K∩L}\max\{\langle \pi^*(c), w\rangle : w \in K \cap L\}max{⟨π∗(c),w⟩:w∈K∩L}, whose value is 111, and w0w_0w0​ is a Slater point for it.

Formalization Note The paper first writes the dual with a full-row-rank matrix MMM whose kernel is L0L_0L0​ and then substitutes z=MTyz = M^{\mathsf T} yz=MTy; the statement here is the second (substituted) form, which is the one the construction uses. L0⊥L_0^\perpL0⊥​ is L.directionᗮ and π∗\pi^*π∗ is LinearMap.adjoint π. The minimum is stated with IsLeast, so attainment is part of the claim. n≥1n \ge 1n≥1 is needed for the same reason as in the previous milestone.

Preamble
import Mathlib
import Definitions.Def_ConeLifts_Factorization_IsConvexBody
import Definitions.Def_ConeLifts_Shared_polar
import Definitions.Def_ConeLifts_Factorization_IsClosedConvexCone
import Definitions.Def_ConeLifts_Shared_dualCone

open scoped InnerProductSpace
Formal statement
namespace ConeLifts.Factorization

/-- Gouveia, Parrilo & Thomas, arXiv:1111.3164v2, Theorem 2.4, proof, p. 4 (the strong-duality
step). Let `C = π(K ∩ L)` with `L = w₀ + L₀` an affine subspace, `π` linear and
`w₀ ∈ int(K)`, and let `c` be an extreme point of `C°`. Then
`1 = min {⟨w₀, z⟩ : z - π*(c) ∈ K*, z ∈ L₀^⊥}` with the minimum attained. Here `L₀` is
`L.direction`, `π*` is the adjoint `LinearMap.adjoint π`, and `K*` is `dualCone K`.
`1 ≤ n` is the paper's implicit full-dimensionality of `C`. -/
theorem slater_certificate {n m : ℕ} (hn : 1 ≤ n)
    (C : Set (EuclideanSpace ℝ (Fin n))) (hC : IsConvexBody C)
    (K : Set (EuclideanSpace ℝ (Fin m))) (hK : IsClosedConvexCone K)
    (hKint : (interior K).Nonempty)
    (L : AffineSubspace ℝ (EuclideanSpace ℝ (Fin m)))
    (π : EuclideanSpace ℝ (Fin m) →ₗ[ℝ] EuclideanSpace ℝ (Fin n))
    (w₀ : EuclideanSpace ℝ (Fin m)) (hw₀L : w₀ ∈ L) (hw₀K : w₀ ∈ interior K)
    (hCπ : C = π '' (K ∩ (L : Set (EuclideanSpace ℝ (Fin m)))))
    (c : EuclideanSpace ℝ (Fin n)) (hc : c ∈ Set.extremePoints ℝ (ConeLifts.Shared.polar C)) :
    IsLeast {t : ℝ | ∃ z ∈ L.directionᗮ,
        z - LinearMap.adjoint π c ∈ ConeLifts.Shared.dualCone K ∧ t = ⟪w₀, z⟫_ℝ} 1 := by sorry

end ConeLifts.Factorization
Source
Gouveia, Parrilo & Thomas, Lifts of Convex Sets and Cone Factorizations, arXiv:1111.3164v2, p. 4, Theorem 2.4 (proof)
Human review
  • Endorsed by Shuze Chen · Sep 27, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Sep 27, 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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me