Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 8.4 — validity of both mixing inequalities

Proved
PochetWolsey.Mixing.mixing_cuts_valid

by Shuze Chen · Oct 29, 2026 · Mathlib 0df444a (Lean v4.33.1)

convex-hullmixed-integer-programmingmixing-setp2o-gran-per-chapterp2o-plan-bookp2o-v1

Let b∈RKb\in\mathbb R^Kb∈RK and let TTT be a nonempty coordinate subset with an enumeration i1,…,iqi_1,\ldots,i_qi1​,…,iq​ such that fi1≤⋯≤fiqf_{i_1}\le\cdots\le f_{i_q}fi1​​≤⋯≤fiq​​, where fk=bk−⌊bk⌋f_k=b_k-\lfloor b_k\rfloorfk​=bk​−⌊bk​⌋ and fi0=0f_{i_0}=0fi0​​=0. Every (s,y)∈XKMIX(b)(s,y)\in X^{MIX}_K(b)(s,y)∈XKMIX​(b) satisfies both inequalities

s≥∑τ=1q(fiτ−fiτ−1)(⌊biτ⌋+1−yiτ),s\ge\sum_{\tau=1}^q(f_{i_\tau}-f_{i_{\tau-1}})(\lfloor b_{i_\tau}\rfloor+1-y_{i_\tau}),s≥τ=1∑q​(fiτ​​−fiτ−1​​)(⌊biτ​​⌋+1−yiτ​​), s≥∑τ=1q(fiτ−fiτ−1)(⌊biτ⌋+1−yiτ)+(1−fiq)(⌊bi1⌋−yi1).s\ge\sum_{\tau=1}^q(f_{i_\tau}-f_{i_{\tau-1}})(\lfloor b_{i_\tau}\rfloor+1-y_{i_\tau})+(1-f_{i_q})(\lfloor b_{i_1}\rfloor-y_{i_1}).s≥τ=1∑q​(fiτ​​−fiτ−1​​)(⌊biτ​​⌋+1−yiτ​​)+(1−fiq​​)(⌊bi1​​⌋−yi1​​).

These are companion validity results on the same mixing model.

Formalization Note An injective map from Fin (t+1) enumerates the nonempty subset, with weak monotonicity preserving ties. Nonemptiness is implicit in the printed use of first and last elements. The bounds 0≤fk<10\le f_k<10≤fk​<1 follow from the definition of fractional part and are not additional assumptions. The singleton case is included.

Preamble
import Mathlib
import Definitions.Def_PochetWolsey_Mixing_Cuts
Formal statement
namespace PochetWolsey.Mixing

theorem mixing_cuts_valid {K t : ℕ} (b : Fin K → ℝ)
    (i : Fin (t + 1) → Fin K) (hi : Function.Injective i)
    (horder : Monotone (fun j => frac (b (i j)))) :
    ∀ p ∈ mixingSet b, cutRhs b i p.2 ≤ p.1 ∧ wrapCutRhs b i p.2 ≤ p.1 := by sorry

end PochetWolsey.Mixing
Source
Pochet and Wolsey, Production Planning by Mixed Integer Programming, Springer 2006, p. 242 (PDF 256), Proposition 8.4, Eqs. (8.7)–(8.8)
Read-back

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

Inputs and hypotheses.

  • Natural numbers K,tK, tK,t.
  • b∈RKb \in \mathbb{R}^Kb∈RK.
  • A map i:{0,…,t}→{0,…,K−1}i : \{0,\dots,t\} \to \{0,\dots,K-1\}i:{0,…,t}→{0,…,K−1} that is injective.
  • The values fj:=f(bij)=bij−⌊bij⌋f_j := f(b_{i_j}) = b_{i_j} - \lfloor b_{i_j}\rfloorfj​:=f(bij​​)=bij​​−⌊bij​​⌋ are nondecreasing: f0≤f1≤⋯≤ftf_0 \le f_1 \le \dots \le f_tf0​≤f1​≤⋯≤ft​.

Conclusion. Every (s,y)(s,y)(s,y) in

Xbmix={(s,y):s≥0, yk∈Z, bk≤s+yk ∀k}X^{\mathrm{mix}}_b = \{(s,y) : s\ge 0,\ y_k\in\mathbb{Z},\ b_k\le s+y_k\ \forall k\}Xbmix​={(s,y):s≥0, yk​∈Z, bk​≤s+yk​ ∀k}

satisfies both cutb,i(y)≤s\mathrm{cut}_{b,i}(y) \le scutb,i​(y)≤s and wcutb,i(y)≤s\mathrm{wcut}_{b,i}(y) \le swcutb,i​(y)≤s. Written out with βj:=bij\beta_j := b_{i_j}βj​:=bij​​, these two inequalities are

f0(⌊β0⌋+1−yi0)+∑j=0t−1(fj+1−fj)(⌊βj+1⌋+1−yij+1)≤sf_0(\lfloor\beta_0\rfloor+1-y_{i_0}) + \sum_{j=0}^{t-1}(f_{j+1}-f_j)(\lfloor\beta_{j+1}\rfloor+1-y_{i_{j+1}}) \le sf0​(⌊β0​⌋+1−yi0​​)+j=0∑t−1​(fj+1​−fj​)(⌊βj+1​⌋+1−yij+1​​)≤s

and

[the same left-hand side]+(1−ft)(⌊β0⌋−yi0)≤s.\Big[\text{the same left-hand side}\Big] + (1-f_t)(\lfloor\beta_0\rfloor - y_{i_0}) \le s .[the same left-hand side]+(1−ft​)(⌊β0​⌋−yi0​​)≤s.

Degenerate cases.

  • Injectivity forces t+1≤Kt+1 \le Kt+1≤K.
  • If K=0K = 0K=0, no such iii exists and the statement is vacuous.
  • If t=0t = 0t=0, monotonicity is automatic and the sum is empty. The claims become:
    • f0(⌊β0⌋+1−yi0)≤sf_0(\lfloor\beta_0\rfloor+1-y_{i_0}) \le sf0​(⌊β0​⌋+1−yi0​​)≤s;
    • f0(⌊β0⌋+1−yi0)+(1−f0)(⌊β0⌋−yi0)≤sf_0(\lfloor\beta_0\rfloor+1-y_{i_0}) + (1-f_0)(\lfloor\beta_0\rfloor-y_{i_0}) \le sf0​(⌊β0​⌋+1−yi0​​)+(1−f0​)(⌊β0​⌋−yi0​​)≤s.
  • If every fj=0f_j = 0fj​=0 (all selected bijb_{i_j}bij​​ integers), the first inequality reads 0≤s0 \le s0≤s. The second reads ⌊β0⌋−yi0≤s\lfloor\beta_0\rfloor - y_{i_0} \le s⌊β0​⌋−yi0​​≤s.

There are no junk values.

Human review
  • Endorsed by Community (Bot) · Oct 29, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Shuze Chen · Oct 29, 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