roseking's picture
Add lean source and binary files (part 41)
97c3405 verified
Raw History Blame Contribute Delete
7.79 kB
import OAI.Combinatorics.CliqueFree.VertexSplitting
import OAI.Combinatorics.CliqueFree.Peeling
namespace OAI
noncomputable section
open scoped BigOperators
open Finset
attribute [local instance] Classical.propDecidable
namespace CliqueFreeIndependence.WeightedGraph
universe u
variable {V : Type u} [Fintype V]
omit [Fintype V] in
lemma EdgeProjection.of_le {G H : SimpleGraph V} (hHG : H ≤ G) : EdgeProjection H G id where
map_adj := fun _ _ h ↦ hHG h
inj := fun _ _ _ _ _ _ h1 h2 ↦ ⟨h1,h2⟩
lemma neighborMass_bot (w : V → ℝ) (u : V) : neighborMass (⊥ : SimpleGraph V) w u = 0 := by
simp [neighborMass, neighbors, mass]
lemma triangleMass_bot (w : V → ℝ) : triangleMass (⊥ : SimpleGraph V) w = 0 := by
simp [triangleMass,triangleAt,neighbors]
lemma splittingConstant_nonneg {C : ℝ} (hC : 1 ≤ C) : 0 ≤ splittingConstant C := by
unfold splittingConstant smoothingConstant entropyConstant exceptionalConstant
positivity
/-- One reduction step, allowing an arbitrary finite vertex type of copies. -/
lemma neighborhood_reduction (G : SimpleGraph V) {w : V → ℝ} (hw : ∀ u, 0 < w u)
{C x : ℝ} (hC : 1 ≤ C) (hCross : CrossBound G w C) (hx : 32 ≤ x)
(hε : splittingConstant C * Numerical.lossRatio x ≤ 1)
(hL : ∀ u, neighborMass G w u ≤ Real.exp x) :
∃ (U : Type u) (_ : Fintype U) (J : SimpleGraph U) (π : U → V),
EdgeProjection J G π ∧ (∀ a, neighborMass J (w ∘ π) a ≤ Real.exp (Real.sqrt x)) ∧
(1-splittingConstant C * Numerical.lossRatio x) * triangleMass G w - edgeMass G w / x ≤
triangleMass J (w ∘ π) := by
have hx0 : 0 < x := by linarith
have hw0 := fun v ↦ (hw v).le
have hε0 : 0 ≤ splittingConstant C * Numerical.lossRatio x :=
mul_nonneg (splittingConstant_nonneg hC) (Numerical.lossRatio_nonneg (by linarith))
obtain ⟨H,hHG,hc,hpeel⟩ := peeling G hw (one_div_pos.2 hx0)
have hprojH := EdgeProjection.of_le hHG
have hcH : CrossBound H w C := by
simpa only [Function.comp_id] using hprojH.crossBound hw0 (by linarith) hCross
have hLH : ∀ u, neighborMass H w u ≤ Real.exp x := by
intro u
exact (hprojH.neighborMass_le hw0 u).trans (hL u)
by_cases hne : Nonempty (EdgeState H)
· let := hne
obtain ⟨J,hproj,hJ,hT⟩ := weighted_splitting hw hC hcH hx hc hLH
refine ⟨EdgeState H,inferInstance,J,EdgeState.src,?_,hJ,?_⟩
· exact ⟨fun a b h ↦ hHG (hproj.map_adj a b h),hproj.inj⟩
· change (1-splittingConstant C * Numerical.lossRatio x) * triangleMass H w ≤
triangleMass J (w ∘ EdgeState.src) at hT
have hh := mul_le_mul_of_nonneg_left hpeel (sub_nonneg.2 hε)
have hm := div_nonneg (edgeMass_nonneg G hw0) hx0.le
rw [one_div_mul_eq_div] at hh
nlinarith only [hT, hh, mul_nonneg hε0 hm]
· have hbot : H = ⊥ := by
apply SimpleGraph.ext
funext a b
apply propext
simpa only [SimpleGraph.bot_adj, iff_false] using
(fun hab : H.Adj a b ↦ hne ⟨⟨(a,b),hab⟩⟩)
rw [hbot,triangleMass_bot] at hpeel
refine ⟨V,inferInstance,⊥,id,EdgeProjection.of_le bot_le,?_,?_⟩
· intro a
simp only [Function.comp_id,neighborMass_bot]
exact (Real.exp_pos _).le
· simp only [Function.comp_id,triangleMass_bot]
rw [one_div_mul_eq_div] at hpeel
nlinarith [mul_nonneg hε0 (triangleMass_nonneg G hw0)]
/-- Uniform triangle bound at the reverse-squaring scale. -/
lemma triangle_bound_scale {C X : ℝ} (hC : 1 ≤ C) (hX : 32 ≤ X)
(hsmall : splittingConstant C * Numerical.lossRatio X ≤ 1/4) (n : ℕ) :
∀ {V : Type u} [Fintype V] (G : SimpleGraph V) (w : V → ℝ),
(∀ u, 0 < w u) → CrossBound G w C →
(∀ u, neighborMass G w u ≤ Real.exp (Numerical.scale X n)) →
triangleMass G w ≤ (1+Real.exp X/3) * (2-(1/2:ℝ)^n) * edgeMass G w := by
induction n with
| zero =>
intro V _ G w hw hCross hL
have ht := triangleMass_le_of_neighborMass G (fun v ↦ (hw v).le) hL
simp only [Numerical.scale, pow_zero] at ht ⊢
have he := edgeMass_nonneg G (fun v ↦ (hw v).le)
nlinarith
| succ n ih =>
intro V _ G w hw hCross hL
let x := Numerical.scale X (n+1)
let q : ℝ := (1/2:ℝ)^(n+1)
let ε := splittingConstant C * Numerical.lossRatio x
have hx : 32 ≤ x := hX.trans (Numerical.scale_ge (by linarith) _)
have hq0 : 0 ≤ q := by positivity
have hq1 : q ≤ 1 := pow_le_one₀ (by norm_num) (by norm_num)
have hε0 : 0 ≤ ε := mul_nonneg (splittingConstant_nonneg hC)
(Numerical.lossRatio_nonneg (by linarith))
have hε : ε ≤ q/4 := by
have hh := mul_le_mul_of_nonneg_left (Numerical.lossRatio_scale (by linarith : 16 ≤ X) (n+1))
(splittingConstant_nonneg hC)
have hh2 := mul_le_mul_of_nonneg_left hsmall hq0
dsimp [ε,x,q]
nlinarith only [hh,hh2]
obtain ⟨U,inst,J,π,hproj,hJ,hstep⟩ := neighborhood_reduction G hw hC hCross hx
(by linarith : ε ≤ 1) hL
let := inst
have hsqrt : Real.sqrt x = Numerical.scale X n := by
dsimp [x,Numerical.scale]
exact Real.sqrt_sq (by linarith [Numerical.scale_ge (by linarith : 1 ≤ X) n])
rw [hsqrt] at hJ
have hJcross := hproj.crossBound (fun v ↦ (hw v).le) (by linarith) hCross
have hupper := ih J (w ∘ π) (fun a ↦ hw (π a)) hJcross hJ
have hmass := hproj.edgeMass_le (fun v ↦ (hw v).le)
have hcoeff : 0 ≤ (1+Real.exp X/3)*(2-(1/2:ℝ)^n) := by
apply mul_nonneg (by positivity)
have hp : (1/2:ℝ)^n ≤ 1 := pow_le_one₀ (by norm_num) (by norm_num)
linarith
have hupperG := hupper.trans (mul_le_mul_of_nonneg_left hmass hcoeff)
have hqeq : (1/2:ℝ)^n = 2*q := by dsimp [q]; rw [pow_succ]; ring
rw [hqeq] at hupperG
have hinv : 1/x ≤ q/2 := by
have hi := Numerical.inv_scale (by linarith : 2 ≤ X) (n+1)
have hXi : 1/X ≤ 1/2 := (one_div_le_one_div_of_le (by norm_num) (by linarith))
have hi2 := mul_le_mul_of_nonneg_left hXi hq0
dsimp [x,q] at *
linarith
apply Numerical.recurrence_bound (by linarith [Real.exp_pos X] : 1 ≤ 1+Real.exp X/3)
hq0 hq1 hε0 hε (triangleMass_nonneg G (fun v ↦ (hw v).le))
(edgeMass_nonneg G (fun v ↦ (hw v).le)) hinv
calc
(1-ε)*triangleMass G w - (1/x)*edgeMass G w =
(1-ε)*triangleMass G w - edgeMass G w/x := by ring
_ ≤ triangleMass J (w ∘ π) := hstep
_ ≤ _ := hupperG
/-- The weighted triangle theorem with a size-independent constant. -/
theorem weighted_triangles {C : ℝ} (hC : 1 ≤ C) :
∃ B : ℝ, 0 < B ∧ ∀ {V : Type u} [Fintype V] (G : SimpleGraph V) (w : V → ℝ),
(∀ v, 0 < w v) → CrossBound G w C → triangleMass G w ≤ B * edgeMass G w := by
obtain ⟨X,hX,hsmall⟩ := Numerical.exists_iteration_threshold (splittingConstant C)
refine ⟨2*(1+Real.exp X/3),by positivity,?_⟩
intro V _ G w hw hCross
obtain ⟨n,hn⟩ := exists_nat_ge (mass w (univ : Finset V))
have hL : ∀ v, neighborMass G w v ≤ Real.exp (Numerical.scale X n) := by
intro v
calc
_ ≤ mass w univ := mass_mono (fun v ↦ (hw v).le) (subset_univ _)
_ ≤ n := hn
_ ≤ Numerical.scale X n := by
have hh := Numerical.scale_ge_add (by linarith : 2 ≤ X) n
linarith
_ ≤ Real.exp (Numerical.scale X n) := by linarith [Real.add_one_le_exp (Numerical.scale X n)]
have ht := triangle_bound_scale hC hX hsmall n G w hw hCross hL
have he := edgeMass_nonneg G (fun v ↦ (hw v).le)
have hB : 0 ≤ 1+Real.exp X/3 := by positivity
have hq : 0 ≤ (1/2:ℝ)^n := by positivity
nlinarith [mul_nonneg hB hq, mul_nonneg (mul_nonneg hB hq) he]
end CliqueFreeIndependence.WeightedGraph
end
end OAI