Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(38), p. 1436 — the block set S* = {b : c(b) ≺ c̄, w*_{c̄,t} > w*_{c(b),t} − ∑ y*_{b,d,s}} maximizes the violation of (37)

Proved
OpenPitMIP.Hourglass.most_violated_set

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

mixed-integer-programmingopen-pit-miningp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1separationvalid-inequalities

Consider an instance of the PCPSP-C satisfying the standing assumptions, arbitrary values wc,t∗w^*_{c,t}wc,t∗​ and yb,d,t∗y^*_{b,d,t}yb,d,t∗​ (for instance a solution of the LP relaxation), a cluster cˉ\bar ccˉ, a destination ddd and periods 1≤t1≤t≤T1\le t_1\le t\le T1≤t1​≤t≤T. For a set SSS of blocks let

viol(S)=(q(S∪{cˉ})−∑s=t1tUsd)wcˉ,t∗+∑b∈b(R)qb∑s=t1tyb,d,s∗−∑b∈S∪{cˉ}qb(wc(b),t∗−∑s=t1tyb,d,s∗)\mathrm{viol}(S)=\Bigl(q(S\cup\{\bar c\})-\sum_{s=t_1}^{t}U^d_s\Bigr)w^*_{\bar c,t}+\sum_{b\in b(R)}q_b\sum_{s=t_1}^{t}y^*_{b,d,s}-\sum_{b\in S\cup\{\bar c\}}q_b\Bigl(w^*_{c(b),t}-\sum_{s=t_1}^{t}y^*_{b,d,s}\Bigr)viol(S)=(q(S∪{cˉ})−s=t1​∑t​Usd​)wcˉ,t∗​+b∈b(R)∑​qb​s=t1​∑t​yb,d,s∗​−b∈S∪{cˉ}∑​qb​(wc(b),t∗​−s=t1​∑t​yb,d,s∗​)

be the violation of the hourglass inequality (37), with R=rcl(cˉ)∖{cˉ}R=rcl(\bar c)\setminus\{\bar c\}R=rcl(cˉ)∖{cˉ} and S∪{cˉ}S\cup\{\bar c\}S∪{cˉ} the blocks of SSS together with those of cˉ\bar ccˉ. Then the set

S∗={b∈B: c(b)≺cˉ, wcˉ,t∗>wc(b),t∗−∑s=t1tyb,d,s∗}(38)S^*=\Bigl\{b\in\mathcal B:\ c(b)\prec\bar c,\ w^*_{\bar c,t}>w^*_{c(b),t}-\sum_{s=t_1}^{t}y^*_{b,d,s}\Bigr\}\tag{38}S∗={b∈B: c(b)≺cˉ, wcˉ,t∗​>wc(b),t∗​−s=t1​∑t​yb,d,s∗​}(38)

satisfies S∗⊆b(cl(cˉ)∖{cˉ})S^*\subseteq b(cl(\bar c)\setminus\{\bar c\})S∗⊆b(cl(cˉ)∖{cˉ}), and viol(S)≤viol(S∗)\mathrm{viol}(S)\le\mathrm{viol}(S^*)viol(S)≤viol(S∗) for every S⊆b(cl(cˉ)∖{cˉ})S\subseteq b(cl(\bar c)\setminus\{\bar c\})S⊆b(cl(cˉ)∖{cˉ}).

This makes the separation of the most violated hourglass cut, for fixed cˉ,d,t1,t\bar c,d,t_1,tcˉ,d,t1​,t, a simple pass over the blocks.

Formalization Note The page says S∗S^*S∗ "maximizes the violation"; with ties the maximizer is not unique, so the statement is that S∗S^*S∗ attains the maximum. The maximum is over all S⊆b(cl(cˉ)∖{cˉ})S\subseteq b(cl(\bar c)\setminus\{\bar c\})S⊆b(cl(cˉ)∖{cˉ}), as on the page; the side condition q(S)≥∑s=t1tUsdq(S)\ge\sum_{s=t_1}^{t}U^d_sq(S)≥∑s=t1​t​Usd​ of Theorem 8 is not imposed, and S∗S^*S∗ need not satisfy it. No feasibility is assumed of (w∗,y∗)(w^*,y^*)(w∗,y∗).

Preamble
import Mathlib
import Definitions.Def_OpenPitMIP_Hourglass_Setting
Formal statement
namespace OpenPitMIP.Hourglass

open Classical in
/-- Display (38), p. 1436: the most violated hourglass cut. For arbitrary values `w*` and `y*`
(for instance a solution of the LP relaxation), a cluster `c̄`, a destination `d` and periods
`t₁ ≤ t`, let `viol S` be the left-hand side minus the right-hand side of (37) at `(w*, y*)`.
The block set `S* = {b : c(b) ≺ c̄, w*_{c̄,t} > w*_{c(b),t} − ∑_{s=t₁}^{t} y*_{b,d,s}}` is admissible,
`S* ⊆ b(cl(c̄) \ {c̄})`, and attains the maximum of `viol` over all `S ⊆ b(cl(c̄) \ {c̄})`. -/
theorem most_violated_set {B D C : Type} [Fintype B] [Fintype D] [Fintype C]
    [DecidableEq B] [DecidableEq C] {T m : ℕ}
    (I : PCPSPC B D C T m) (hI : I.Standing)
    (w : C → Fin T → ℝ) (y : B → D → Fin T → ℝ) (cbar : C) (d : D)
    (t₁ t : Fin T) (ht : t₁ ≤ t) :
    let viol : Finset B → ℝ := fun S =>
      (I.qBlocks (S ∪ I.blocksOf {cbar}) - I.Ucum d t₁ t) * w cbar t
        + ∑ b ∈ I.blocksOf (I.rcl cbar \ {cbar}), I.q b * PCPSPC.ySum y b d t₁ t
        - ∑ b ∈ S ∪ I.blocksOf {cbar}, I.q b * (w (I.clu b) t - PCPSPC.ySum y b d t₁ t)
    let Sstar : Finset B := Finset.univ.filter
      (fun b => I.cprec (I.clu b) cbar ∧ w (I.clu b) t - PCPSPC.ySum y b d t₁ t < w cbar t)
    Sstar ⊆ I.blocksOf (I.cl cbar \ {cbar}) ∧
      ∀ S ⊆ I.blocksOf (I.cl cbar \ {cbar}), viol S ≤ viol Sstar := by sorry

end OpenPitMIP.Hourglass
Source
Oper. Res. 68(5), display (38), §5.2.3, p. 1436
Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 9, 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