Lemma 7.5 (used in the proof of Theorem 1.3) — exact optimum and decoding
OpenIntervalMaxCutEPTAS.Scheme.Supplement.lemma_7_5complexity-theorygraph-theorymax-cutp2o-batch-stoc27sp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a simple graph with vertices, edges and , and let . Then
Moreover, every target cut , with read from the two core copies, satisfies
The exact optimum and the universal deficit inequality are one item. All arithmetic in the signed deficit is real, with the positive denominator . This quantitative connection permits an approximation gap to pass through the reduction.
Differences from the source:
- encoding: finite vertices and one-based copy and repetition indices → Fin n, Fin 2, and Fin p; dependent private indices; reason: these are bijective renamings, with exactly the source number of vertices and no padding
- encoding: the signed deficit divided by 4p → subtraction and division in ℝ after casting finite cut counts; reason: this avoids truncated natural arithmetic, and p ≥ 1 ensures the denominator is positive
Preamble
import Mathlib import Definitions.Def_IntervalMaxCutEPTAS_Scheme_Supplement_Gadget set_option autoImplicit false
Formal statement
namespace IntervalMaxCutEPTAS.Scheme.Supplement
/-- Lemma 7.5, PDF 27 (printed 26). The subtraction and division in the deficit inequality
are performed in ℝ, with the positive denominator 4p. -/
theorem lemma_7_5 {n : ℕ} (Q : SimpleGraph (Fin n)) [DecidableRel Q.Adj]
(p : ℕ) (hm : 1 ≤ Q.edgeFinset.card) (hp : 1 ≤ p) :
mc (target Q p) = 4 * p * (2 * Q.edgeFinset.card + n + mc Q) + n ^ 2 ∧
∀ S : Finset (TargetVertex Q p),
max (cutCount Q (sourceSide Q p 0 S) : ℝ) (cutCount Q (sourceSide Q p 1 S) : ℝ) ≥
(mc Q : ℝ) - ((mc (target Q p) : ℝ) - (cutCount (target Q p) S : ℝ)) / (4 * (p : ℝ)) := by sorry
end IntervalMaxCutEPTAS.Scheme.Supplement
Source
Morteza Alimi, Hani Ahmadzadeh, STOC 2027 submission #636 (October 2026), Lemma 7.5, PDF p. 27 (printed p. 26)