Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 7.2.1 — existence and uniqueness of the primal–dual central path

Proved
MatousekLP.InteriorPoint.central_path_exists_unique

by mikedeng1 · Oct 3, 2026 · Mathlib 0df444a (Lean v4.33.1)

central-pathinterior-point-methodslinear-programmingp2o-batch-b23bp2o-gran-per-chapterp2o-plan-bookp2o-v1

Let AAA be a real m×nm\times nm×n matrix of rank mmm, b∈Rmb\in\mathbb{R}^mb∈Rm, c∈Rnc\in\mathbb{R}^nc∈Rn, and consider the linear program (7.2), maximize cTxc^{T}xcTx subject to Ax=bAx=bAx=b, x≥0x\ge 0x≥0, and its dual (7.5), minimize bTyb^{T}ybTy subject to ATy≥cA^{T}y\ge cATy≥c. Suppose (7.2) has a feasible solution x~>0\tilde x>0x~>0 and (7.5) has a feasible solution y~\tilde yy~​ whose slack vector s~=ATy~−c\tilde s=A^{T}\tilde y-cs~=ATy~​−c satisfies s~>0\tilde s>0s~>0; that is, both programs have an interior feasible point.

Then for every μ>0\mu>0μ>0 the system

Ax=b,ATy−s=c,(s1x1,…,snxn)=μ1,x,s≥0(7.4)Ax=b,\qquad A^{T}y-s=c,\qquad (s_1x_1,\dots,s_nx_n)=\mu\mathbf 1,\qquad x,s\ge 0 \tag{7.4}Ax=b,ATy−s=c,(s1​x1​,…,sn​xn​)=μ1,x,s≥0(7.4)

in the unknowns x,s∈Rnx,s\in\mathbb{R}^nx,s∈Rn, y∈Rmy\in\mathbb{R}^my∈Rm has exactly one solution x∗(μ),y∗(μ),s∗(μ)x^*(\mu),y^*(\mu),s^*(\mu)x∗(μ),y∗(μ),s∗(μ), and x∗(μ)x^*(\mu)x∗(μ) is the unique maximizer of the barrier function fμ(x)=cTx+μ∑j=1nln⁡xjf_\mu(x)=c^{T}x+\mu\sum_{j=1}^n\ln x_jfμ​(x)=cTx+μ∑j=1n​lnxj​ subject to Ax=bAx=bAx=b and x>0x>0x>0.

The set {(x∗(μ),y∗(μ),s∗(μ)):μ>0}\{(x^*(\mu),y^*(\mu),s^*(\mu)):\mu>0\}{(x∗(μ),y∗(μ),s∗(μ)):μ>0} is the primal–dual central path of (7.2), the curve that path-following interior point methods track.

Formalization Note Indices are Fin n and Fin m. Uniqueness of the triple is stated as: one solution exists, and every solution of (7.4) equals it. "Unique maximizer" means Ax∗=bAx^*=bAx∗=b, x∗>0x^*>0x∗>0, and fμ(x)<fμ(x∗)f_\mu(x)<f_\mu(x^*)fμ​(x)<fμ​(x∗) for every other xxx with Ax=bAx=bAx=b, x>0x>0x>0; the barrier is never evaluated at a point with a nonpositive coordinate in this comparison. The rank hypothesis is the standing assumption of (7.2).

Preamble
import Mathlib
import Definitions.Def_MatousekLP_InteriorPoint_CentralPath
Formal statement
namespace MatousekLP.InteriorPoint

open Matrix

/-- Lemma 7.2.1 (Matoušek & Gärtner, p. 121). Let `A` be an `m × n` matrix of rank `m`.
Suppose the linear program (7.2) `maximize cᵀx subject to Ax = b, x ≥ 0` has a feasible
solution `x̃ > 0` and the dual (7.5) `minimize bᵀy subject to Aᵀy ≥ c` has a feasible solution
`ỹ` whose slack vector `s̃ = Aᵀỹ − c` satisfies `s̃ > 0`. Then for every `μ > 0` the system
(7.4) has a unique solution `x*(μ), y*(μ), s*(μ)`, and `x*(μ)` is the unique maximizer of
`f_μ` subject to `Ax = b` and `x > 0`. -/
theorem central_path_exists_unique {m n : ℕ} (A : Matrix (Fin m) (Fin n) ℝ) (b : Fin m → ℝ)
    (c : Fin n → ℝ) (hrank : A.rank = m) (xt : Fin n → ℝ) (hxt : IsPrimalInterior A b xt)
    (yt : Fin m → ℝ) (hyt : IsDualInterior A c yt) :
    ∀ μ : ℝ, 0 < μ →
      ∃ (x : Fin n → ℝ) (y : Fin m → ℝ) (s : Fin n → ℝ),
        CentralPathSystem A b c μ x y s ∧
        (∀ (x' : Fin n → ℝ) (y' : Fin m → ℝ) (s' : Fin n → ℝ),
          CentralPathSystem A b c μ x' y' s' → x' = x ∧ y' = y ∧ s' = s) ∧
        IsUniqueBarrierMaximizer A b c μ x := by sorry

end MatousekLP.InteriorPoint
Source
Matoušek & Gärtner, Understanding and Using Linear Programming, Springer 2007, §7.2, p. 121, Lemma 7.2.1 (with (7.2) p. 119, (7.4) and (7.5) p. 120)
Human review
  • Endorsed by Shuze Chen · Oct 5, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 5, 2026

    Confirmed by the mission captain (proposal self-audit).

  • Endorsed by Community (Bot) · Oct 20, 2026

    Confirmed by the moderator at approval.

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