Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Overlapping net windows and pairwise consistency for exact regional recovery

Definition
StableMSR_WindowCSP

by lreyzin · Nov 21, 2026 · Mathlib 0df444a (Lean v4.33.1)

algorithmsclusteringformal-verificationgraph-theory

A net point chooses a candidate partition of its radius-4R window. Only clusters that touch its owned points participate in compatibility checks, and each nonempty cluster is charged to the owner of its least point. A scenario encodes a radius pair and one region label per point. The two program contracts separate geometric scenario generation from the uniform exact-count optimization of these pairwise choices.

Definition code
import Definitions.Def_StableMSR_LocalBallTrials
import Definitions.Def_StableMSR_RegionalSeeds
import Definitions.Def_StableMSR_Recovery_Supplement_Partitions
import Definitions.Def_RobertsonSeymour1986_GM5_TreewidthLE
set_option autoImplicit false
open scoped Classical
open StableMSR.Recovery StableMSR.Recovery.Compiler StableMSR.Recovery.OracleQueries
open StableMSR.Recovery.BallCoverCatalog StableMSR.Recovery.FewBallRows
open StableMSR.Recovery.RegionalSeeds StableMSR.Recovery.Supplement
namespace StableMSR.Recovery.WindowCSP
noncomputable def budget (e : ℝ) : ℕ := 2*(ownerBound e+1)
noncomputable def window {n : ℕ} (d : Point n→Point n→ℝ) (W : Cluster n) (R : ℝ) (a : Point n) :=
  W∩ball d a (4*R)
noncomputable def anchor {n : ℕ} (d : Point n→Point n→ℝ) (W : Cluster n) (R : ℝ) (i : Point n) : Point n :=
  Supplement.representative ((greedyNet d R W).filter (fun a=>d a i≤2*R)) i
def touches {n : ℕ} (owner : Point n→Point n) (a : Point n) (C : Cluster n) : Prop :=
  ∃i∈C,owner i=a
def holds {n : ℕ} (owner : Point n→Point n) (a : Point n) (C : Cluster n) : Prop :=
  ∃i∈C,owner i=a ∧ ∀j∈C,i≤ j
noncomputable def accepted {n : ℕ} (owner : Point n→Point n) (a : Point n) (P : Partition n) :=
  P.filter (touches owner a)
noncomputable def owned {n : ℕ} (owner : Point n→Point n) (a : Point n) (P : Partition n) :=
  P.filter (holds owner a)
abbrev Pick (n : ℕ) := Point n→Partition n
noncomputable def compatible {n : ℕ} (owner : Point n→Point n) (a b : Point n) (P Q : Partition n) : Prop :=
  ∀C∈accepted owner a P,∀D∈accepted owner b Q,(C∩D).Nonempty → C=D
noncomputable def valid (x : World) (W : Cluster x.n) (R : ℝ) (h : ℕ) (P : Pick x.n) : Prop :=
  (∀a∈greedyNet x.dist R W,P a∈traceMenu x h (window x.dist W R a)) ∧
  ∀a b : ↥(greedyNet x.dist R W),(interactionGraph x.dist R W).Adj a b →
    compatible (anchor x.dist W R) a b (P a) (P b)
noncomputable def output (x : World) (W : Cluster x.n) (R : ℝ) (P : Pick x.n) : Partition x.n :=
  (greedyNet x.dist R W).biUnion (fun a=>owned (anchor x.dist W R) a (P a))
noncomputable def number (x : World) (W : Cluster x.n) (R : ℝ) (P : Pick x.n) : ℕ :=
  ∑a ∈ greedyNet x.dist R W,(owned (anchor x.dist W R) a (P a)).card
noncomputable def price (x : World) (W : Cluster x.n) (R : ℝ) (P : Pick x.n) : ℝ :=
  ∑a ∈ greedyNet x.dist R W,cost x.dist (owned (anchor x.dist W R) a (P a))
def sceneBase (x : World) (s : ℕ) := 3+s*(x.n+2)
noncomputable def scale (x : World) (s : ℕ) :=
  MinSumFPT.oracleAnswer x.dist (x.mem (sceneBase x s)) (x.mem (sceneBase x s+1))
def label (x : World) (s : ℕ) (i : Point x.n) := x.mem (sceneBase x s+2+i.val)
noncomputable def regions (x : World) (s : ℕ) : Partition x.n :=
  Finset.univ.image (fun i=>Finset.univ.filter (fun j=>label x s j=label x s i))
def sceneSpec (e : ℚ) (a b : ℕ) (cmd : Cmd) (n k : ℕ) (d : Point n→Point n→ℝ) : Prop :=
  ∃s z H, Executes d cmd (initialStore n k) s z ∧ z≤ a*n^(b*parameterEnvelope (e:ℝ)) ∧
    H≤ a*n^(b*parameterEnvelope (e:ℝ)) ∧
    let x:=worldOf k H d s
    Ready x s ∧ ∀O,Stable d k (1+(e:ℝ)) O → fewBound (e:ℝ)<(largeClusters d O (e:ℝ)).card →
      ∃u< H,0< scale x u ∧ scale x u=largestRadius d O ∧
        ∀W∈regions x u,Aligned O W ∧ diameter d W≤ densityScale d O (e:ℝ)
def searchSpec (e : ℚ) (c D b : ℕ) (cmd : Cmd) : Prop :=
  ∀(x : World) (s : Store),1≤ x.k → x.k< x.n → IsMetricArray x.dist → Ready x s →
    ∃t z H, Executes x.dist cmd s t z ∧ z≤ c*x.cap^D*x.n^(b*parameterEnvelope (e:ℝ)) ∧
      H≤ c*x.cap^D*x.n^(b*parameterEnvelope (e:ℝ)) ∧
      let y:=worldOf x.k H x.dist t
      Ready y t ∧ endpoints y ∧ ∀u< x.rows,0< scale x u →
        (∀W∈regions x u,RobertsonSeymour1986.GM5.TreewidthLE (interactionGraph x.dist (scale x u) W)
          (widthBound (e:ℝ))) →
        ∀P : Cluster x.n→Pick x.n,(∀W∈regions x u,valid x W (scale x u) (budget (e:ℝ)) (P W)) →
        (∑W ∈ regions x u,number x W (scale x u) (P W))=x.k →
        ∃r< H,good y r ∧ score y r≤∑W ∈ regions x u,price x W (scale x u) (P W)
def originalRows (x : World) := (x.mem 0).toNat
def tableSlots (x : World) := originalRows x+1
noncomputable def unaryField (x : World) (r j : ℕ) (b : Bool) : ℤ :=
  let m:=tableSlots x
  if r=0 then (m:ℤ)
  else if r≤ m+m then ∑i ∈ Finset.range x.rows,
    if i< originalRows x ∧ m+(i+2)=r then RegionalBallDP.field x i j b else 0
  else if r≤ m+m+m then (x.k:ℤ)
  else if b then ∑q ∈ Finset.range x.rows, if m+m+m+q=r then (q:ℤ) else 0
  else 1
end StableMSR.Recovery.WindowCSP
Source
Exact Sum-of-Radii Clustering under Any Constant Stability, arXiv:2611.03547, Lemmas 3.3, 3.8, 3.9(iii), 4.4, Proposition 4.2, Algorithm 1, and Appendix A.4–A.6. Alternative pairwise formulation using bounded-overlap window traces in the mission’s real-distance model.

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