Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§5 — ψ=(ρ−n)log⁡(xTs)−∑jlog⁡xjsjxTs/n+nlog⁡n≥(ρ−n)log⁡(xTs)+nlog⁡n\psi = (\rho-n)\log(x^Ts) - \sum_j \log\frac{x_js_j}{x^Ts/n} + n\log n \ge (\rho-n)\log(x^Ts) + n\log nψ=(ρ−n)log(xTs)−∑j​logxTs/nxj​sj​​+nlogn≥(ρ−n)log(xTs)+nlogn

Proved
AdaptiveStepIPM.Potential.psi_rewrite_lower_bound

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

interior-point-methodslinear-programmingp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1potential-reduction

Let n≥1n \ge 1n≥1, ρ∈R\rho \in \mathbb Rρ∈R, and x,s∈Rnx, s \in \mathbb R^nx,s∈Rn with x>0x > 0x>0, s>0s > 0s>0. Let ψ(x,s)=ρlog⁡(xTs)−∑j=1nlog⁡(xjsj)\psi(x,s) = \rho\log(x^Ts) - \sum_{j=1}^n\log(x_js_j)ψ(x,s)=ρlog(xTs)−∑j=1n​log(xj​sj​) be the primal–dual potential (1). Then:

  1. the potential can be rewritten as
ψ(x,s)=(ρ−n)log⁡(xTs)−∑j=1nlog⁡(xjsjxTs/n)+nlog⁡n;\psi(x,s) = (\rho - n)\log(x^Ts) - \sum_{j=1}^n \log\Bigl(\frac{x_j s_j}{x^Ts/n}\Bigr) + n\log n ;ψ(x,s)=(ρ−n)log(xTs)−j=1∑n​log(xTs/nxj​sj​​)+nlogn;
  1. the sum ∑jlog⁡(xjsj/(xTs/n))\sum_{j}\log\bigl(x_js_j/(x^Ts/n)\bigr)∑j​log(xj​sj​/(xTs/n)) is nonpositive (arithmetic–geometric mean inequality), so
ψ(x,s)≥(ρ−n)log⁡(xTs)+nlog⁡n;\psi(x,s) \ge (\rho - n)\log(x^Ts) + n\log n ;ψ(x,s)≥(ρ−n)log(xTs)+nlogn;
  1. consequently, if ρ>n\rho > nρ>n, t≥0t \ge 0t≥0 and ψ(x,s)≤−(ρ−n)t+nlog⁡n\psi(x,s) \le -(\rho-n)t + n\log nψ(x,s)≤−(ρ−n)t+nlogn, then xTs≤2−tx^Ts \le 2^{-t}xTs≤2−t.

The lower bound turns a decrease of the potential into a decrease of the duality gap: once ψ\psiψ has been driven down to −(ρ−n)t+nlog⁡n-(\rho-n)t + n\log n−(ρ−n)t+nlogn, precision ttt has been reached.

Formalization Note ψ\psiψ is the published LinearOptimization.interiorPointPotential ρ x s, which agrees with (1) for x,s>0x, s > 0x,s>0. All logarithms are natural; the third item uses e−t≤2−te^{-t} \le 2^{-t}e−t≤2−t for t≥0t \ge 0t≥0.

Preamble
import Mathlib
import Definitions.Def_LinearOptimization_InteriorPointPotential
import Definitions.Def_AdaptiveStepIPM_Potential_Neighborhood

open Matrix
Formal statement
namespace AdaptiveStepIPM.Potential

/-- **§5, p. 11.** For `x, s > 0` the potential (1) can be written as
`ψ(x, s) = (ρ − n) log(xᵀs) − Σⱼ log(x_j s_j/(xᵀs/n)) + n log n`; the sum is nonpositive
(arithmetic–geometric mean inequality), so `ψ(x, s) ≥ (ρ − n) log(xᵀs) + n log n`; hence,
for `ρ > n` and `t ≥ 0`, `ψ(x, s) ≤ −(ρ − n)t + n log n` implies `xᵀs ≤ 2^{−t}`. -/
theorem psi_rewrite_lower_bound {n : ℕ} (hn : 1 ≤ n) (ρ : ℝ) (x s : Fin n → ℝ)
    (hx : ∀ j, 0 < x j) (hs : ∀ j, 0 < s j) :
    LinearOptimization.interiorPointPotential ρ x s =
        (ρ - n) * Real.log (x ⬝ᵥ s) -
          ∑ j, Real.log (x j * s j / ((x ⬝ᵥ s) / n)) + n * Real.log n ∧
      (ρ - n) * Real.log (x ⬝ᵥ s) + n * Real.log n ≤
        LinearOptimization.interiorPointPotential ρ x s ∧
      ∀ t : ℝ, (n : ℝ) < ρ → 0 ≤ t →
        LinearOptimization.interiorPointPotential ρ x s ≤ -((ρ - n) * t) + n * Real.log n →
        x ⬝ᵥ s ≤ (2 : ℝ) ^ (-t) := by sorry

end AdaptiveStepIPM.Potential
Source
Mizuno, Todd, Ye, On adaptive-step primal-dual interior-point algorithms for linear programming, Cornell ORIE Technical Report No. 944 (1990, rev. 1991), p. 11, §5 (displays before Algorithm 3)
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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