Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

A Lagrangian saddle point is primal optimal

Disproved
VectorSpaceOpt.lagrangian_saddle_sufficient

by wenxinzhang · Aug 25, 2026 · Mathlib c5ea003 (Lean v4.30.0)

lagrangianoptimality-certificatesaddle-pointstrong-duality

Let PPP be a closed convex cone, x0∈Ωx_0\in\Omegax0​∈Ω, and let z0∗z_0^*z0∗​ be dual-positive. Assume (x0,z0∗)(x_0,z_0^*)(x0​,z0∗​) is a saddle point in the orientation

L(x0,z∗)≤L(x0,z0∗)≤L(x,z0∗)L(x_0,z^*)\le L(x_0,z_0^*)\le L(x,z_0^*)L(x0​,z∗)≤L(x0​,z0∗​)≤L(x,z0∗​)

for every dual-positive z∗z^*z∗ and every x∈Ωx\in\Omegax∈Ω. Then G(x0)≤P0G(x_0)\le_P0G(x0​)≤P​0, and x0x_0x0​ minimizes fff over all feasible points in Ω\OmegaΩ. The theorem does not assume feasibility separately; closed-cone dual separation makes it a conclusion. This source milestone packages a certificate that is useful independently of convexity or Slater hypotheses and fixes the maximizing multiplier/minimizing primal orientation used by the capstone. It can serve as a standalone verification theorem for candidate solutions.

Preamble
import Definitions.Def_VectorSpaceOpt_coneLE
import Definitions.Def_VectorSpaceOpt_dualPositive
import Definitions.Def_VectorSpaceOpt_lagrangian
Formal statement
namespace VectorSpaceOpt

/-- Luenberger, Chapter 8, §8.4, Theorem 2. -/
theorem lagrangian_saddle_sufficient
    {X Z : Type*}
    [NormedAddCommGroup Z] [NormedSpace ℝ Z]
    (P : ConvexCone ℝ Z) (Ω : Set X) (f : X → ℝ) (G : X → Z)
    (x₀ : X) (zstar₀ : Z →L[ℝ] ℝ)
    (hP : IsClosed (P : Set Z)) (hx₀ : x₀ ∈ Ω)
    (hzstar₀ : dualPositive P zstar₀)
    (hleft : ∀ zstar : Z →L[ℝ] ℝ, dualPositive P zstar →
      lagrangian f G zstar x₀ ≤ lagrangian f G zstar₀ x₀)
    (hright : ∀ x ∈ Ω,
      lagrangian f G zstar₀ x₀ ≤ lagrangian f G zstar₀ x) :
    coneLE P (G x₀) 0 ∧
      ∀ x ∈ Ω, coneLE P (G x) 0 → f x₀ ≤ f x := by sorry

end VectorSpaceOpt
Source
David G. Luenberger, Optimization by Vector Space Methods, Wiley, 1969, public scan https://sites.science.oregonstate.edu/~show/old/142_Luenberger.pdf, Chapter 8, §8.4, Theorem 2, printed p. 221; physical PDF p. 241.
Read-back

What the Lean code literally says, in plain math · gpt-5

Let XXX be an arbitrary type and ZZZ a real normed vector space. For a real convex cone PPP in ZZZ, set Ω⊆XΩ⊆XΩ⊆X, functions f:X→Rf : X → ℝf:X→R and G:X→ZG : X → ZG:X→Z, point x0∈Xx₀∈Xx0​∈X, and continuous real-linear functional z0∗z₀^*z0∗​, assume PPP is closed, x0∈Ωx₀∈Ωx0​∈Ω, and z0∗z₀^*z0∗​ is nonnegative on PPP. Assume also that for every continuous real-linear z∗z^*z∗ nonnegative on PPP, f(x0)+z∗(G(x0))≤f(x0)+z0∗(G(x0))f(x₀)+z^*(G(x₀))≤f(x₀)+z₀^*(G(x₀))f(x0​)+z∗(G(x0​))≤f(x0​)+z0∗​(G(x0​)), and that for every x∈Ωx∈Ωx∈Ω, f(x0)+z0∗(G(x0))≤f(x)+z0∗(G(x))f(x₀)+z₀^*(G(x₀))≤f(x)+z₀^*(G(x))f(x0​)+z0∗​(G(x0​))≤f(x)+z0∗​(G(x)). Then −G(x0)∈P-G(x₀)∈P−G(x0​)∈P, and for every x∈Ωx∈Ωx∈Ω with −G(x)∈P-G(x)∈P−G(x)∈P, f(x0)≤f(x)f(x₀)≤f(x)f(x0​)≤f(x). Thus the first saddle inequality quantifies over all dual-positive multipliers and the second over all points of ΩΩΩ; no convexity of ΩΩΩ, fff, or GGG, and no algebraic or topological structure on XXX, is assumed.

Human review
  • Endorsed by Shuze Chen · Aug 26, 2026

  • Endorsed by wenxinzhang · Aug 26, 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