Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 5.7 — Weight deficits across pieces

Open
GreedyRegionGrowing.Layouts.Supplement.lemma_5_7

by mikedeng1 · Nov 15, 2026 · Mathlib 0df444a (Lean v4.33.1)

combinatoricsgraph-algorithmsoptimizationp2o-batch-stoc27sp2o-gran-per-chapterp2o-plan-paperp2o-v1

Assume (5.1), (5.3) and all Section 5 standing assumptions. In every complete operational run of Algorithm 2, let X_i be the pieces in removal order and Γ(S)=log(e|A|/A(S)). Any maximizing center and any event-free open-annulus radius meeting the exact Line-5 inequality are allowed. Then:

Γ(Xi)≤2Γ(Xj)(i<j).\Gamma(X_i)\le2\Gamma(X_j)\qquad(i<j).Γ(Xi​)≤2Γ(Xj​)(i<j).

This is a supporting result for Theorem 1.5.

Differences from the source:

  • encoding: finite labelled graphs, ordered iterations and polynomial-time algorithms → Fin n carriers, zero-based trace indices, fixed binary codes and deterministic TM2 polynomial time; reason: labels and iteration shifts change no source instance; the codes include all vertices and rational data, and the machine condition records the full runtime claim
  • encoding: Section 5 edge lengths and shortest-path distances → symmetric nonnegative rational extensions to all pairs and a minimum over paths of at most n vertices; reason: only edge lengths enter valid paths; zero extension gives every source instance, and a shortest simple path exists under the connected nonnegative-length standing assumptions
Preamble
import Mathlib
import Definitions.Def_GreedyRegionGrowing_Layouts_Foundation

noncomputable section
open Finset QAlgorithms.Complexity
Formal statement
namespace GreedyRegionGrowing.Layouts.Supplement
attribute [local instance] Classical.propDecidable

/-- §3.1 and Appendix D.3: exactly one commodity for each unordered pair. -/
private def pairs {n : ℕ} (U : Finset (Fin n)) := (U ×ˢ U).filter (fun p => p.1 < p.2)

private def IsDemand {n : ℕ} (U : Finset (Fin n)) (D : Fin n → Fin n → ℝ) : Prop :=
  (∀ u v, 0 ≤ D u v ∧ D u v = D v u) ∧ (∀ u, D u u = 0) ∧
  ∀ u v, u ∉ U ∨ v ∉ U → D u v = 0

private def Respects {n : ℕ} (U : Finset (Fin n)) (A : Fin n → ℚ)
    (D : Fin n → Fin n → ℝ) : Prop :=
  IsDemand U D ∧ ∀ u ∈ U, ∑ v ∈ U, D u v ≤ (A u : ℝ)

/-- Nonnegative flows in both edge orientations, with per-commodity conservation and total load.
No edge outside the induced component can carry a commodity. -/
private def Routable {n : ℕ} (G : SimpleGraph (Fin n)) (U : Finset (Fin n))
    (D : Fin n → Fin n → ℝ) (q : ℝ) : Prop :=
  ∃ f : (Fin n × Fin n) → Fin n → Fin n → ℝ,
    (∀ p x y, 0 ≤ f p x y) ∧
    (∀ p x y, ¬ (x ∈ U ∧ y ∈ U ∧ G.Adj x y) → f p x y = 0) ∧
    (∀ p ∈ pairs U, ∀ x ∈ U,
      (∑ y ∈ U, (f p x y - f p y x)) =
        if x = p.1 then D p.1 p.2 else if x = p.2 then -D p.1 p.2 else 0) ∧
    ∀ e ∈ edges G U, ∑ p ∈ pairs U, (f p e.1 e.2 + f p e.2 e.1) ≤ q

private def FlowExpanding {n : ℕ} (G : SimpleGraph (Fin n)) (U : Finset (Fin n))
    (A : Fin n → ℚ) (phi : ℚ) : Prop :=
  ∀ D, Respects U A D → Routable G U D (1/(phi : ℝ))

private def degreeWeight {n : ℕ} (G : SimpleGraph (Fin n)) (U : Finset (Fin n)) (u : Fin n) : ℚ :=
  (U.filter (fun v => G.Adj u v)).card

private def FlowExpander {n : ℕ} (G : SimpleGraph (Fin n)) (U : Finset (Fin n)) (phi : ℚ) :=
  FlowExpanding G U (degreeWeight G U) phi

private def weight {n : ℕ} (A : Fin n → ℚ) (U : Finset (Fin n)) : ℚ := ∑ v ∈ U, A v

private def productDemand {n : ℕ} (A : Fin n → ℚ) : Fin n → Fin n → ℝ := fun u v =>
  if u = v then 0 else (A u : ℝ) * (A v : ℝ) / (weight A univ : ℝ)

/-- Delete exactly the sorted unordered pairs of C. -/
private def remaining {n : ℕ} (G : SimpleGraph (Fin n)) (C : Finset (Fin n × Fin n)) :
    SimpleGraph (Fin n) where
  Adj u v := G.Adj u v ∧ (u,v) ∉ C ∧ (v,u) ∉ C
  symm := ⟨by intro u v h; exact ⟨G.symm.symm u v h.1, h.2.2, h.2.1⟩⟩
  loopless := ⟨by intro u h; exact G.loopless.irrefl u h.1⟩

private def nodeWeightCode {n : ℕ} (A : Fin n → ℚ) : Str :=
  (List.finRange n).flatMap (fun v => ratCode (A v))

private def edgeSetCode {n : ℕ} (C : Finset (Fin n × Fin n)) : Str :=
  listCode (fun e => natCode e.1.val ++ natCode e.2.val) (((List.finRange n).flatMap (fun u => (List.finRange n).map (fun v => (u,v)))).filter (fun e => decide (e ∈ C)))

/-- Output tag true certifies expansion; tag false carries the selected vertex set. -/
private def certificateCode {n : ℕ} (S : Option (Finset (Fin n))) : Str :=
  match S with
  | none => [true]
  | some X => [false] ++ setCode X

private def loglog (x : ℝ) : ℝ := max 1 (Real.log (Real.log x))

/-- Lengths of simple candidate walks, enumerated on the finite vertex set. -/
private def pathLength {n : ℕ} (ell : Fin n → Fin n → ℚ) : List (Fin n) → ℚ
  | [] => 0
  | [_] => 0
  | u::v::xs => ell u v + pathLength ell (v::xs)

private def pathLengths {n : ℕ} (G : SimpleGraph (Fin n)) (ell : Fin n → Fin n → ℚ)
    (u v : Fin n) : Finset ℚ :=
  ((range (n+1)).biUnion (fun k =>
    (univ : Finset (Fin k → Fin n)).image (fun f => List.ofFn f))).filter
      (fun p => p.head? = some u ∧ p.getLast? = some v ∧ p.IsChain G.Adj) |>.image (pathLength ell)

/-- Nonnegative shortest-path metric. The minimum is over paths with at most n vertices;
zero is used only if no path exists, a case excluded by LengthInstance. -/
private def shortest {n : ℕ} (G : SimpleGraph (Fin n)) (ell : Fin n → Fin n → ℚ)
    (u v : Fin n) : ℚ :=
  let P := pathLengths G ell u v
  if h : P.Nonempty then P.min' h else 0

/-- Section 5 standing assumptions and (5.1); no rounding conclusion is assumed. -/
private def LengthInstance {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (phi : ℚ) (ell : Fin n → Fin n → ℚ) : Prop :=
  G.Connected ∧ (∀ v, 1 ≤ A v) ∧ 0 < phi ∧
  (∀ u v, 0 ≤ ell u v ∧ ell u v = ell v u) ∧
  (∑ e ∈ edges G univ, ell e.1 e.2) ≤ 1 ∧
  (1/(2*(phi : ℝ))) < ∑ p ∈ pairs (univ : Finset (Fin n)),
    productDemand A p.1 p.2 * (shortest G ell p.1 p.2 : ℝ)

private def radiusR {n : ℕ} (A : Fin n → ℚ) (phi : ℚ) : ℚ := 1/(16*phi*weight A univ)

private def SmallBalls {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (phi : ℚ) (ell : Fin n → Fin n → ℚ) : Prop :=
  ∀ u, weight A (ball (shortest G ell) univ u (radiusR A phi : ℝ)) < weight A univ / 3

private def weightedVolume {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (ell : Fin n → Fin n → ℚ) (U : Finset (Fin n)) (u : Fin n) (r : ℝ) : ℝ :=
  if r ≤ 0 then 0 else
  let B := ball (shortest G ell) U u r
  (weight A B : ℝ)/(weight A univ : ℝ) +
    (∑ e ∈ edges G B, (ell e.1 e.2 : ℝ)) +
    ∑ e ∈ cut G U B, if e.1 ∈ B then r-(shortest G ell u e.1 : ℝ)
      else r-(shortest G ell u e.2 : ℝ)

private def deficitF {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (ell : Fin n → Fin n → ℚ) (U : Finset (Fin n)) (u : Fin n) (r : ℝ) : ℝ :=
  Real.log (2*Real.exp 1 / weightedVolume G A ell U u r)

private def gamma {n : ℕ} (A : Fin n → ℚ) (S : Finset (Fin n)) : ℝ :=
  Real.log (Real.exp 1 * (weight A univ : ℝ)/(weight A S : ℝ))

private def kappa {n : ℕ} (A : Fin n → ℚ) : ℕ :=
  2 * (Int.ceil (Real.logb 2 (Real.log (2*Real.exp 1*(weight A univ : ℝ))))).toNat + 1

private def annulus {n : ℕ} (A : Fin n → ℚ) (phi : ℚ) (j : ℕ) : ℝ :=
  (radiusR A phi : ℝ) - j*(radiusR A phi : ℝ)/(2*kappa A)

private def Stable {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ) (phi : ℚ)
    (ell : Fin n → Fin n → ℚ) (U : Finset (Fin n)) (u : Fin n) (j : ℕ) : Prop :=
  deficitF G A ell U u (annulus A phi j) ≤ 2*deficitF G A ell U u (annulus A phi (j-1)) ∧
  gamma A (ball (shortest G ell) U u (annulus A phi j)) ≤
    2*gamma A (ball (shortest G ell) U u (annulus A phi (j-1)))

private structure WeightedFrame (n : ℕ) where
  U : Finset (Fin n)
  pointer : Fin n → ℕ
  center : Fin n
  radius : ℝ

namespace WeightedFrame
variable {n : ℕ}
private def X (f : WeightedFrame n) (G : SimpleGraph (Fin n)) (ell : Fin n → Fin n → ℚ) :=
  ball (shortest G ell) f.U f.center f.radius

private def V (f : WeightedFrame n) (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (ell : Fin n → Fin n → ℚ) := weightedVolume G A ell f.U f.center f.radius

/-- The while-loop increases a pointer through precisely the unstable annuli; no bound κ is imposed. -/
private def Operational (f : WeightedFrame n) (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (phi : ℚ) (ell : Fin n → Fin n → ℚ) (old : Fin n → ℕ) : Prop :=
  f.center ∈ f.U ∧
  (∀ u ∈ f.U, old u ≤ f.pointer u ∧ Stable G A phi ell f.U u (f.pointer u) ∧
    ∀ j, old u ≤ j → j < f.pointer u → ¬ Stable G A phi ell f.U u j) ∧
  (∀ u ∈ f.U,
    weight A (ball (shortest G ell) f.U u (annulus A phi (f.pointer u-1))) ≤
    weight A (ball (shortest G ell) f.U f.center (annulus A phi (f.pointer f.center-1)))) ∧
  (let a := annulus A phi (f.pointer f.center)
   let b := annulus A phi (f.pointer f.center-1)
   a < f.radius ∧ f.radius < b ∧
   f.radius ∉ (f.U.image (fun v => (shortest G ell f.center v : ℝ))) ∧
   ((cut G f.U (f.X G ell)).card : ℝ) ≤ 4/(b-a)*f.V G A ell*
     deficitF G A ell f.U f.center f.radius)
end WeightedFrame

private def WeightedRunFrom {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ) (phi : ℚ)
    (ell : Fin n → Fin n → ℚ) : Finset (Fin n) → (Fin n → ℕ) → List (WeightedFrame n) → Prop
  | _, _, [] => True
  | U, old, f::fs => f.U = U ∧ f.Operational G A phi ell old ∧
    WeightedRunFrom G A phi ell (U \ f.X G ell) f.pointer fs

private def WeightedRun {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ) (phi : ℚ)
    (ell : Fin n → Fin n → ℚ) (fs : List (WeightedFrame n)) :=
  WeightedRunFrom G A phi ell univ (fun _ => 1) fs

private def WeightedComplete {n : ℕ} (G : SimpleGraph (Fin n)) (ell : Fin n → Fin n → ℚ)
    (fs : List (WeightedFrame n)) : Prop :=
  match fs.getLast? with
  | none => n = 0
  | some f => f.U \ f.X G ell = ∅

/-- Line 7 of Algorithm 2: every minimizing piece is an allowed returned set. -/
private def WeightedOutput {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (ell : Fin n → Fin n → ℚ) (fs : List (WeightedFrame n)) (S : Finset (Fin n)) : Prop :=
  (∃ f ∈ fs, S = f.X G ell) ∧
  ∀ f ∈ fs, ((cut G univ S).card : ℝ)/((weight A S : ℝ)*gamma A S) ≤
    ((cut G univ (f.X G ell)).card : ℝ)/((weight A (f.X G ell) : ℝ)*gamma A (f.X G ell))

/-- Coded promise input for Lemmas 5.3 and 5.4, lengths included. -/
private structure RoundingInput where
  n : ℕ
  G : SimpleGraph (Fin n)
  A : Fin n → ℚ
  phi : ℚ
  ell : Fin n → Fin n → ℚ
  valid : LengthInstance G A phi ell

private def roundingCode (I : RoundingInput) : Str :=
  graphCode I.G ++ nodeWeightCode I.A ++ ratCode I.phi ++
    (List.finRange I.n).flatMap (fun u => (List.finRange I.n).flatMap (fun v => ratCode (I.ell u v)))

private structure HeavyInput extends RoundingInput where
  Y : Finset (Fin n)
  heavy : weight A univ / 3 ≤ weight A Y
  diameter : ∀ u ∈ Y, ∀ v ∈ Y, shortest G ell u v ≤ 2*radiusR A phi


/-- Aditya Anand, STOC 2027 submission #506 (October 2026), lemma 5 7. -/
theorem lemma_5_7 {n : ℕ} (G : SimpleGraph (Fin n)) (A : Fin n → ℚ)
    (phi : ℚ) (ell : Fin n → Fin n → ℚ) (hI : LengthInstance G A phi ell)
    (hsmall : SmallBalls G A phi ell) (fs : List (WeightedFrame n))
    (hr : WeightedRun G A phi ell fs) (hc : WeightedComplete G ell fs) :
  ∀ i j : Fin fs.length, i < j → gamma A ((fs.get i).X G ell) ≤ 2*gamma A ((fs.get j).X G ell) := by sorry

end GreedyRegionGrowing.Layouts.Supplement
Source
Aditya Anand, Greedy Region Growing: Improved Approximations for Cutwidth, Pathwidth, Flow-Expander Decomposition, and More, STOC 2027 submission #506 (October 2026). Citation basis: published (canonical held submission); 46 pages; printed page = PDF page throughout (offset 0); no arXiv version. Lemma 5.7, p. 28 (PDF p. 28).

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