Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 1, p. 6 — (6a)–(6c) is an ideal formulation of gr(ReLU∘f; [L, U])

Proved
ReluMIP.Ideal.proposition_1

by mikedeng1 · Oct 8, 2026 · Mathlib 0df444a (Lean v4.33.1)

ideal-formulationmixed-integer-programmingneural-networkp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1relu

Let f(x)=w⋅x+bf(x)=w\cdot x+bf(x)=w⋅x+b be an affine function with w∈Rηw\in\mathbb R^\etaw∈Rη and b∈Rb\in\mathbb Rb∈R, over the input domain [L,U][L,U][L,U] with Li<UiL_i<U_iLi​<Ui​ for every iii, and assume strict activity: M−(f)<0<M+(f)M^-(f)<0<M^+(f)M−(f)<0<M+(f), where M+(f)=w⋅U˘+bM^+(f)=w\cdot\breve U+bM+(f)=w⋅U˘+b and M−(f)=w⋅L˘+bM^-(f)=w\cdot\breve L+bM−(f)=w⋅L˘+b are the maximum and minimum of fff on [L,U][L,U][L,U]. Then

y≥w⋅x+b(6a)y≤∑i∈Iwi(xi−L˘i(1−z))+(b+∑i∉IwiU˘i)z∀I⊆supp⁡(w)(6b)(x,y,z)∈[L,U]×R≥0×{0,1}(6c)\begin{aligned} &y\ge w\cdot x+b &&(6a)\\ &y\le\sum_{i\in I}w_i\bigl(x_i-\breve L_i(1-z)\bigr)+\Bigl(b+\sum_{i\notin I}w_i\breve U_i\Bigr)z\quad\forall I\subseteq\operatorname{supp}(w) &&(6b)\\ &(x,y,z)\in[L,U]\times\mathbb R_{\ge0}\times\{0,1\} &&(6c) \end{aligned}​y≥w⋅x+by≤i∈I∑​wi​(xi​−L˘i​(1−z))+(b+i∈/I∑​wi​U˘i​)z∀I⊆supp(w)(x,y,z)∈[L,U]×R≥0​×{0,1}​​(6a)(6b)(6c)​

is an ideal formulation of gr⁡(ReLU∘f;[L,U])\operatorname{gr}(\mathrm{ReLU}\circ f;[L,U])gr(ReLU∘f;[L,U]). That is:

  1. (x,y)∈gr⁡(ReLU∘f;[L,U])(x,y)\in\operatorname{gr}(\mathrm{ReLU}\circ f;[L,U])(x,y)∈gr(ReLU∘f;[L,U]) if and only if (x,y,z)(x,y,z)(x,y,z) satisfies (6a)–(6c) for some z∈{0,1}z\in\{0,1\}z∈{0,1};
  2. every extreme point of the LP relaxation of (6) (with z∈[0,1]z\in[0,1]z∈[0,1] in place of z∈{0,1}z\in\{0,1\}z∈{0,1}) has z∈{0,1}z\in\{0,1\}z∈{0,1}.

The formulation uses only the original variables (x,y)(x,y)(x,y) and one binary variable, with exponentially many inequalities (6b); it is the strongest possible tightening of the big-M formulation (3).

Formalization Note The two hypotheses Li<UiL_i<U_iLi​<Ui​ and strict activity are the standing assumptions of §1.3 of the paper. "Ideal" follows the paper's definition (p. 3): the extreme points of the LP relaxation are integral; only zzz is an integer variable, so integrality concerns zzz alone.

Preamble
import Mathlib
import Definitions.Def_ReluMIP_Ideal_Setting
Formal statement
namespace ReluMIP.Ideal

/-- Proposition 1, p. 6 (arXiv:1811.08359v2; proof App. A.1, pp. 14–15): for `f(x) = w · x + b`
over `[L, U]` (with the standing assumptions of §1.3: `Lᵢ < Uᵢ` and strict activity), (6a)–(6c) is
an ideal formulation of `gr(ReLU ∘ f; [L, U])`: together with `z ∈ {0, 1}` its LP relaxation
formulates the graph, and every extreme point of the LP relaxation has `z ∈ {0, 1}`. -/
theorem proposition_1 {η : ℕ} (w : Fin η → ℝ) (b : ℝ) (L U : Fin η → ℝ)
    (hLU : ∀ i, L i < U i) (hSA : StrictActivity w b L U) :
    IsFormulationFor (relax6 w b L U) (reluGraph w b L U) ∧ IsIdeal (relax6 w b L U) := by sorry

end ReluMIP.Ideal
Source
arXiv:1811.08359v2, Proposition 1, p. 6 (proof App. A.1, pp. 14–15); standing assumptions §1.3, p. 4; definition of ideal §1.1, p. 3
Human review
  • Endorsed by Shuze Chen · Oct 8, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 8, 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