Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2.13 — r/b ≤ k ≤ r for an inequality with defect r and N-index k

Proved
LovaszSchrijver.Defect.defect_div_le_NIndex_le_defect

by mikedeng1 · Sep 27, 2026 · Mathlib 0df444a (Lean v4.33.1)

graph-theorylift-and-projectp2o-batch-p100bp2o-gran-per-chapterp2o-plan-paperp2o-v1stable-set-polytope

Let G=(V,E)G = (V,E)G=(V,E) be a finite graph with no isolated nodes, and let aTx≤ba^{\mathsf T}x \le baTx≤b be an inequality with integer coefficients a∈Z+Va \in \mathbb Z_+^Va∈Z+V​, b∈Z+b \in \mathbb Z_+b∈Z+​, valid for STAB(G)\mathrm{STAB}(G)STAB(G), with defect

r=2max⁡{aTx−b:x∈FRAC(G)}≥0r = 2\max\{a^{\mathsf T}x - b : x \in \mathrm{FRAC}(G)\} \ge 0r=2max{aTx−b:x∈FRAC(G)}≥0

and N-index kkk (the least ttt such that aTx≤ba^{\mathsf T}x \le baTx≤b is valid for Nt(G)N^t(G)Nt(G)). Then

rb≤k≤r,\frac{r}{b} \le k \le r,br​≤k≤r,

stated with the denominator cleared as

r≤k bandk≤r.r \le k\,b \qquad\text{and}\qquad k \le r.r≤kbandk≤r.

The theorem shows that the number of rounds of the Lovász–Schrijver NNN operator needed to derive a stable set inequality is controlled by how far the fractional stable set polytope violates it: an inequality violated by at most r/2r/2r/2 on FRAC(G)\mathrm{FRAC}(G)FRAC(G) is valid after rrr rounds, and cannot be valid before r/br/br/b rounds.

Formalization Note

  1. The lower bound is written r≤kbr \le k br≤kb rather than r/b≤kr/b \le kr/b≤k. For b>0b > 0b>0 the two are equivalent; for b=0b = 0b=0 validity for STAB(G)\mathrm{STAB}(G)STAB(G) forces a=0a = 0a=0 and r=0r = 0r=0, and the cleared form avoids Lean's convention r/0=0r/0 = 0r/0=0.
  2. The hypothesis r≥0r \ge 0r≥0 is added. The paper's proof starts "If r=0r = 0r=0 we have nothing to prove, so suppose that r>0r > 0r>0", presuming r≥0r \ge 0r≥0; without it the upper bound fails (for x1≤2x_1 \le 2x1​≤2 on a single edge, r=−2r = -2r=−2 while k=0k = 0k=0). It holds whenever b=max⁡{aTx:x∈STAB(G)}b = \max\{a^{\mathsf T}x : x \in \mathrm{STAB}(G)\}b=max{aTx:x∈STAB(G)}, because STAB(G)⊆FRAC(G)\mathrm{STAB}(G) \subseteq \mathrm{FRAC}(G)STAB(G)⊆FRAC(G).
  3. The defect and the N-index are given as values rrr, kkk with IsGreatest and IsLeast hypotheses, never as a supremum or infimum. Coefficients are natural numbers cast to R\mathbb RR.
Preamble
import Mathlib
import Definitions.Def_LovaszSchrijver_Defect_Index
Formal statement
namespace LovaszSchrijver.Defect

theorem defect_div_le_NIndex_le_defect {V : Type} [Fintype V] [DecidableEq V]
    (G : SimpleGraph V) [DecidableRel G.Adj] (hG : ∀ v, ∃ w, G.Adj v w)
    (a : V → ℕ) (b : ℕ) (hvalid : Valid (STAB G) (fun i => (a i : ℝ)) b)
    (r : ℝ) (hr : IsDefect G (fun i => (a i : ℝ)) b r) (hr0 : 0 ≤ r)
    (k : ℕ) (hk : IsNIndex G (fun i => (a i : ℝ)) b k) :
    r ≤ k * b ∧ (k : ℝ) ≤ r := by sorry

end LovaszSchrijver.Defect
Source
Lovász and Schrijver, Cones of matrices and set-functions and 0–1 optimization, SIAM J. Optim. 1(2) (1991), p. 182, Theorem 2.13
Read-back

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

Throughout, VVV is any type with finitely many elements and decidable equality. GGG is a simple graph on the vertex set VVV (no loops, undirected), and its adjacency relation is decidable. Four notions come from the imported definitions bundle: STAB\mathrm{STAB}STAB, Valid\mathrm{Valid}Valid, IsDefect\mathrm{IsDefect}IsDefect and IsNIndex\mathrm{IsNIndex}IsNIndex. Their definitions are not part of the code given here, so this read-back can only name them. What they actually require has to be checked in that bundle.

The statement has these inputs and hypotheses:

  • No isolated vertices. Every vertex has at least one neighbour:
∀v∈V ∃w∈V: v∼Gw.\forall v \in V\ \exists w \in V:\ v \sim_G w.∀v∈V ∃w∈V: v∼G​w.
  • An inequality with natural-number data. There is a coefficient vector a:V→Na : V \to \mathbb{N}a:V→N and a right-hand side b∈Nb \in \mathbb{N}b∈N. The vector enters every predicate below as the real vector (ai)i∈V(a_i)_{i \in V}(ai​)i∈V​, with each aia_iai​ viewed as a real number.
  • Validity. The pair ((ai)i∈V, b)\big((a_i)_{i\in V},\, b\big)((ai​)i∈V​,b) satisfies Valid(STAB(G),a,b)\mathrm{Valid}(\mathrm{STAB}(G), a, b)Valid(STAB(G),a,b). Here STAB(G)\mathrm{STAB}(G)STAB(G) is an object built from GGG by the bundle's definition STAB\mathrm{STAB}STAB.
  • A defect value. rrr is a real number with IsDefect(G,a,b,r)\mathrm{IsDefect}(G, a, b, r)IsDefect(G,a,b,r) and r≥0r \ge 0r≥0.
  • An N-index value. kkk is a natural number with IsNIndex(G,a,b,k)\mathrm{IsNIndex}(G, a, b, k)IsNIndex(G,a,b,k).

Under all these hypotheses, the theorem concludes both inequalities, computed in the real numbers:

r  ≤  k⋅bandk  ≤  r.r \;\le\; k \cdot b \qquad\text{and}\qquad k \;\le\; r.r≤k⋅bandk≤r.

The conclusion holds for every real rrr that satisfies the defect predicate with r≥0r \ge 0r≥0, and for every natural kkk that satisfies the N-index predicate. The code does not say that either predicate picks out a unique value. The two inequalities together imply k≤kbk \le k bk≤kb, and also k≤rk \le rk≤r with r≥0r \ge 0r≥0. Both inequalities are non-strict (≤\le≤).

Degenerate cases. If VVV is empty, the no-isolated-vertex hypothesis holds vacuously. The graph then has no vertices and aaa is the empty vector. The conclusion still requires k≤r≤kbk \le r \le k bk≤r≤kb, provided the three bundle predicates can hold for that empty data.

If VVV has exactly one element, the no-isolated-vertex hypothesis cannot hold, because a simple graph has no loops. The theorem is therefore vacuous for one-vertex VVV.

If b=0b = 0b=0, the conclusion becomes r≤0r \le 0r≤0 and k≤rk \le rk≤r. With r≥0r \ge 0r≥0, this forces r=0r = 0r=0 and k=0k = 0k=0. If k=0k = 0k=0 (with any bbb), the conclusion likewise forces r=0r = 0r=0.

The statement uses no division, no subtraction and no suprema or integrals, so no default values from Lean's total functions enter it directly. Any such default values would sit inside the unseen definitions of STAB\mathrm{STAB}STAB, Valid\mathrm{Valid}Valid, IsDefect\mathrm{IsDefect}IsDefect and IsNIndex\mathrm{IsNIndex}IsNIndex.

Whether those three predicates can hold at the same time for a given GGG, aaa and bbb cannot be determined from this code. If they never can, the theorem is vacuous.

Human review
  • Endorsed by Shuze Chen · Sep 27, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Sep 27, 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