openAI-math-dataset / lean /OAI /Geometry /Cannon /Boundary /VisualSynchronousPrefix.lean
roseking's picture
Add lean source and binary files (part 66)
8e4d533 verified
Raw History Blame Contribute Delete
17.8 kB
import Mathlib
import OAI.Geometry.Cannon.Boundary.WordBall
namespace OAI
section
section
section
noncomputable section
section
section
open Set Topology
namespace CannonRelease
open CannonGraph CannonVisual Filter Topology
universe v
variable {G : Type v} [Group G] [TopologicalSpace G] [DiscreteTopology G]
variable (D : CayleyData G) {δ : ℕ} (hδ : Thin D.graph δ)
local instance visualSynchronousPrefixSetoid : Setoid (BasedRay D) := raySetoid D
theorem visual_synchronous_prefix (C : ℝ) (_hC : 0 ≤ C) :
∃ B : ℕ, ∀ (r s : BasedRay D) (j : ℕ),
boundaryDist D hδ (Quotient.mk' r) (Quotient.mk' s) ≤ C*visualScale δ j →
D.graph.dist (r.val j) (s.val j) ≤ B := by
have hε := visualExponent_pos δ
have hexp : Tendsto (fun n : ℕ => Real.exp (visualExponent δ*(n:ℝ))) atTop atTop :=
Real.tendsto_exp_atTop.comp (tendsto_natCast_atTop_atTop.const_mul_atTop hε)
obtain ⟨A,hA⟩ := (hexp.eventually (eventually_gt_atTop (2*C))).exists
refine ⟨2*(A+(8*δ+6))+4*δ+2,fun r s j hj => ?_⟩
have hc := (visualMetric_comparison D hδ r s).2
have hp : (preDist (visualExponent δ) r s : ℝ) <
Real.exp (visualExponent δ*(A:ℝ)-2*visualExponent δ*j) := by
calc
_ ≤ 2*boundaryDist D hδ (Quotient.mk' r) (Quotient.mk' s) := hc
_ ≤ 2*(C*visualScale δ j) := mul_le_mul_of_nonneg_left hj (by norm_num)
_ < Real.exp (visualExponent δ*(A:ℝ))*visualScale δ j := by
nlinarith only [mul_lt_mul_of_pos_right hA (visualScale_pos δ j)]
_ = _ := by rw [visualScale,← Real.exp_add]; congr 1; push_cast; ring
have hlim : Tendsto (fun n => (decay (visualExponent δ) r s n : ℝ)) atTop
(𝓝 (preDist (visualExponent δ) r s : ℝ)) :=
NNReal.continuous_coe.tendsto _ |>.comp
(tendsto_decay D.connected hε.le r s)
obtain ⟨n,hn,hdec⟩ := ((eventually_ge_atTop j).and (hlim.eventually_lt_const hp)).exists
change Real.exp (-visualExponent δ*(product2 (G:=D.graph) 1 (r.val n) (s.val n):ℝ)) < _ at hdec
have hprod : 2*j ≤ product2 (G:=D.graph) 1 (r.val n) (s.val n)+A := by
have he := Real.exp_lt_exp.mp hdec
have he' : (2*j:ℕ) ≤ product2 (G:=D.graph) 1 (r.val n) (s.val n)+A := by
have hh : ((2*j:ℕ):ℝ) ≤ ((product2 (G:=D.graph) 1 (r.val n) (s.val n)+A:ℕ):ℝ) := by
push_cast
nlinarith only [he,hε]
exact_mod_cast hh
exact he'
have hpref := (product2_prefix D.connected hδ r s hn).2
have hr : D.graph.dist 1 (r.val j) = j := by
simpa only [r.property.1,Nat.dist_zero_left] using r.property.2 0 j
dsimp only [CayleyData.graph] at *
have hv := vertex_close_to_ray D.connected hδ s (r.val j) n (A+(8*δ+6))
(by simpa only [hr] using hn) (by rw [hr]; omega)
simpa only [hr] using hv
theorem separated_ray_products {e : ℝ} (he : 0 < e) :
∃ C : ℕ, ∀ (r s : BasedRay D),
e ≤ boundaryDist D hδ (Quotient.mk' r) (Quotient.mk' s) →
∀ n, product2 (G:=D.graph) 1 (r.val n) (s.val n) ≤ C := by
have hε := visualExponent_pos δ
have ht : Tendsto (fun n : ℕ => Real.exp (-visualExponent δ*(n:ℝ))) atTop (𝓝 0) := by
apply Real.tendsto_exp_atBot.comp
simpa only [neg_mul,Function.comp_def] using tendsto_neg_atTop_atBot.comp
(tendsto_natCast_atTop_atTop.const_mul_atTop hε)
obtain ⟨C,hC⟩ := (ht.eventually_lt_const he).exists
refine ⟨C,fun r s hsep n => ?_⟩
have hc := (visualMetric_comparison D hδ r s).1
have hd := preDist_le_decay (visualExponent δ) r s n
have hd' : (preDist (visualExponent δ) r s : ℝ) ≤
Real.exp (-visualExponent δ*(product2 (G:=D.graph) 1 (r.val n) (s.val n):ℝ)) := hd
by_contra hn
have hCN : (C:ℝ) ≤ product2 (G:=D.graph) 1 (r.val n) (s.val n) := by exact_mod_cast (lt_of_not_ge hn).le
have hm := Real.exp_le_exp.mpr (mul_le_mul_of_nonpos_left hCN (neg_nonpos.mpr hε.le))
exact (not_lt_of_ge (hsep.trans (hc.trans hd'))) (hm.trans_lt hC)
theorem inverse_puncture_prefix {e : ℝ} (he : 0 < e) :
∃ B : ℕ, ∀ (r z : BasedRay D) (k : ℕ),
e ≤ boundaryDist D hδ (Quotient.mk' r) (Quotient.mk' z) →
D.graph.dist (r.val k)⁻¹
((translateRay D ⟨δ,hδ⟩ (r.val k)⁻¹ z).val k) ≤ B := by
obtain ⟨C,hC⟩ := separated_ray_products D hδ he
refine ⟨2*(C+(8*δ+6))+4*δ+2,fun r z k hz => ?_⟩
let g := (r.val k)⁻¹
let p := translateRay D ⟨δ,hδ⟩ g z
have hk : D.graph.dist 1 (r.val k) = k := by
simpa only [r.property.1,Nat.dist_zero_left] using r.property.2 0 k
have hg : D.graph.dist 1 g = k := by
have hh := D.dist_mul_left g (r.val k) 1
simpa [g,SimpleGraph.dist_comm,hk] using hh
obtain ⟨n,hn,hbig⟩ := ((eventually_ge_atTop k).and
(translated_mixed_product_large D hδ g z (2*D.graph.dist 1 g))).exists
have hrepl := product2_replace_one D.connected hδ 1 g (g*z.val n) (p.val n) hbig
have hpref := (product2_prefix D.connected hδ r z hn).1
have hbound := hC r z hz n
have hcomp := product2_complement D.connected 1 (r.val k) (z.val n)
have hmul : product2 (G:=D.graph) 1 g (g*z.val n) =
product2 (G:=D.graph) (r.val k) 1 (z.val n) := by
have hh := product2_mul D g (r.val k) 1 (z.val n)
simpa [g] using hh
change product2 (G:=D.graph) (r.val k) 1 (z.val n) +
product2 (G:=D.graph) 1 (r.val k) (z.val n) = 2*D.graph.dist 1 (r.val k) at hcomp
rw [hk,← hmul] at hcomp
dsimp only [CayleyData.graph] at *
have hprod : 2*D.graph.dist 1 g ≤ product2 (G:=D.graph) 1 g (p.val n)+(C+(8*δ+6)) := by
dsimp only [CayleyData.graph]
rw [hg]
omega
have hh := vertex_close_to_ray D.connected hδ p g n (C+(8*δ+6)) (hg ▸ hn) hprod
simpa only [hg] using hh
end CannonRelease
namespace CannonRelease
open CannonGraph CannonVisual Filter Topology Finset
universe v
variable {G : Type v} [Group G] [TopologicalSpace G] [DiscreteTopology G]
variable (D : CayleyData G) {δ : ℕ} (hδ : Thin D.graph δ)
local instance visualSynchronousGraphBallSetoid : Setoid (BasedRay D) := raySetoid D
local instance visualSynchronousGraphBallDecidable : ∀ p : Prop, Decidable p :=
Classical.propDecidable
omit [TopologicalSpace G] [DiscreteTopology G] in
lemma uniform_graph_ball_card (B : ℕ) :
∃ N : ℕ, ∀ (x : G) (S : Finset G),
(∀ y ∈ S, D.graph.dist x y ≤ B) → S.card ≤ N := by
classical
let T := (finite_graph_ball D.connected (finite_cayley_neighbors D.generators) 1 B).toFinset
refine ⟨T.card,fun x S hS => ?_⟩
apply card_le_card_of_injOn (fun y => x⁻¹*y) ?_ (by intro a _ b _ hab; exact mul_left_cancel hab)
intro y hy
change x⁻¹*y ∈ (finite_graph_ball D.connected (finite_cayley_neighbors D.generators) 1 B).toFinset
rw [Set.Finite.mem_toFinset]
have hh := D.dist_mul_left x 1 (x⁻¹*y)
simp only [mul_one,mul_inv_cancel_left] at hh
change D.graph.dist 1 (x⁻¹*y) ≤ B
rw [← hh]
exact hS y hy
theorem vertex_assignment_multiplicity {H : ℝ} (hH : 0 < H) :
∃ N : ℕ, ∀ {ι : Type*} (S : Finset ι) (r : ι → BasedRay D) (k : ℕ),
(∀ i ∈ S, ∀ j ∈ S, i ≠ j → H*visualScale δ k ≤
boundaryDist D hδ (Quotient.mk' (r i)) (Quotient.mk' (r j))) →
∀ v : G, (S.filter (fun i => (r i).val k = v)).card ≤ N := by
classical
let _ := visualMetricSpace D hδ
obtain ⟨N,hN⟩ := relative_packing_bound D hδ hH
refine ⟨N,fun {ι} S r k hsep v => ?_⟩
let T := S.filter (fun i => (r i).val k = v)
by_cases hT : T.Nonempty
· obtain ⟨i,hi⟩ := hT
let x : ι → Boundary D := fun j => Quotient.mk' (r j)
have hinj : Set.InjOn x (T : Set ι) := by
intro j hj j' hj' heq
by_contra hne
have hh := hsep j (mem_filter.mp hj).1 j' (mem_filter.mp hj').1 hne
change H*visualScale δ k ≤ dist (x j) (x j') at hh
rw [heq,dist_self] at hh
exact (not_le_of_gt (mul_pos hH (visualScale_pos δ k))) hh
have hh := hN (x i) (visualScale δ k) (visualScale_pos δ k)
(by rw [visualScale]; apply Real.exp_le_one_iff.mpr; exact mul_nonpos_of_nonpos_of_nonneg (neg_nonpos.mpr (visualExponent_pos δ).le) (Nat.cast_nonneg _)) (T.image x) (by
intro y hy
obtain ⟨j,hj,rfl⟩ := mem_image.mp hy
exact boundaryDist_le_of_prefix D hδ (r i) (r j)
((mem_filter.mp hi).2.trans (mem_filter.mp hj).2.symm)) (by
intro y hy y' hy' hne
obtain ⟨j,hj,rfl⟩ := mem_image.mp hy
obtain ⟨j',hj',rfl⟩ := mem_image.mp hy'
exact hsep j (mem_filter.mp hj).1 j' (mem_filter.mp hj').1
(fun he => hne (congrArg x he)))
rwa [card_image_of_injOn hinj] at hh
· change T.card ≤ _
rw [Finset.not_nonempty_iff_eq_empty.mp hT,Finset.card_empty]
exact Nat.zero_le _
theorem puncture_ball_multiplicity {e H C : ℝ} (he : 0 < e) (hH : 0 < H) (hC : 0 ≤ C) :
∃ N : ℕ, ∀ {ι : Type*} (S : Finset ι) (r : ι → BasedRay D) (z : BasedRay D) (k : ℕ),
(∀ i ∈ S, e ≤ boundaryDist D hδ (Quotient.mk' (r i)) (Quotient.mk' z)) →
(∀ i ∈ S, ∀ j ∈ S, i ≠ j → H*visualScale δ k ≤
boundaryDist D hδ (Quotient.mk' (r i)) (Quotient.mk' (r j))) →
∀ b : Boundary D, (S.filter (fun i => boundaryDist D hδ
(boundaryTranslate D ⟨δ,hδ⟩ ((r i).val k)⁻¹ (Quotient.mk' z)) b ≤ C*visualScale δ k)).card ≤ N := by
classical
let _ := visualMetricSpace D hδ
obtain ⟨B,hB⟩ := inverse_puncture_prefix D hδ he
obtain ⟨B',hB'⟩ := visual_synchronous_prefix D hδ (2*C) (by positivity)
obtain ⟨Ng,hNg⟩ := uniform_graph_ball_card D (B+B')
obtain ⟨Nv,hNv⟩ := vertex_assignment_multiplicity D hδ hH
refine ⟨Nv*Ng,fun {ι} S r z k hsep hcent b => ?_⟩
let p : ι → BasedRay D := fun i => translateRay D ⟨δ,hδ⟩ ((r i).val k)⁻¹ z
let T := S.filter (fun i => boundaryDist D hδ (Quotient.mk' (p i)) b ≤ C*visualScale δ k)
change T.card ≤ Nv*Ng
by_cases hT : T.Nonempty
· obtain ⟨i₀,hi₀⟩ := hT
have hnear (i : ι) (hi : i ∈ T) :
D.graph.dist ((r i).val k)⁻¹ ((p i₀).val k) ≤ B+B' := by
have h₁ := hB (r i) z k (hsep i (mem_filter.mp hi).1)
have hpd : boundaryDist D hδ (Quotient.mk' (p i)) (Quotient.mk' (p i₀)) ≤
(2*C)*visualScale δ k := by
have hh := @dist_triangle (Boundary D) _ (Quotient.mk' (p i)) b (Quotient.mk' (p i₀))
rw [@dist_comm (Boundary D) _ b (Quotient.mk' (p i₀))] at hh
change boundaryDist D hδ (Quotient.mk' (p i)) (Quotient.mk' (p i₀)) ≤
boundaryDist D hδ (Quotient.mk' (p i)) b + boundaryDist D hδ (Quotient.mk' (p i₀)) b at hh
linarith only [hh,(mem_filter.mp hi).2,(mem_filter.mp hi₀).2]
have h₂ := hB' (p i) (p i₀) k hpd
have ht := @SimpleGraph.Connected.dist_triangle _ D.graph
((r i).val k)⁻¹ ((p i).val k) ((p i₀).val k) D.connected
exact ht.trans (Nat.add_le_add h₁ h₂)
have hcard := hNg ((p i₀).val k) (T.image (fun i => ((r i).val k)⁻¹)) (by
intro y hy
obtain ⟨i,hi,rfl⟩ := mem_image.mp hy
rw [SimpleGraph.dist_comm]
exact hnear i hi)
have hfib : ∀ v ∈ T.image (fun i => ((r i).val k)⁻¹),
(T.filter (fun i => ((r i).val k)⁻¹ = v)).card ≤ Nv := by
intro v _
have hh := hNv T r k (fun i hi j hj hij => hcent i (mem_filter.mp hi).1
j (mem_filter.mp hj).1 hij) v⁻¹
simpa only [inv_eq_iff_eq_inv] using hh
exact (card_le_mul_card_image T Nv hfib).trans (Nat.mul_le_mul_left Nv hcard)
· rw [Finset.not_nonempty_iff_eq_empty.mp hT,Finset.card_empty]
exact Nat.zero_le _
end CannonRelease
namespace CannonGraph
open SimpleGraph CannonVisual
variable {V : Type*} {Γ : SimpleGraph V} {o : V}
lemma prefix_of_near_ray (hΓ : Γ.Connected) {δ : ℕ} (hδ : Thin Γ δ)
(p : Ray Γ o) (v w : V) (k l B : ℕ) (hl : l ≤ k)
(_hv : Γ.dist o v = k) (hw : Γ.dist o w = l)
(hwv : Γ.dist w v = k-l) (hvp : Γ.dist v (p.val k) ≤ B) :
Γ.dist w (p.val l) ≤ 2*B+4*δ+2 := by
have hp : Γ.dist o (p.val k) = k := by
simpa only [p.property.1,Nat.dist_zero_left] using p.property.2 0 k
have ht := @Connected.dist_triangle _ Γ w v (p.val k) hΓ
have hprod := product2_add_dist hΓ o w (p.val k)
have hlarge : 2*Γ.dist o w ≤ product2 (G:=Γ) o w (p.val k)+B := by
rw [hwv] at ht
rw [hw,hp] at hprod
rw [hw]
omega
have hh := vertex_close_to_ray hΓ hδ p w k B (by simpa only [hw] using hl) hlarge
simpa only [hw] using hh
end CannonGraph
namespace CannonRelease
open CannonGraph CannonVisual Finset
universe v
variable {G : Type v} [Group G] [TopologicalSpace G] [DiscreteTopology G]
variable (D : CayleyData G) {δ : ℕ} (hδ : Thin D.graph δ)
local instance visualSynchronousInverseRaySetoid : Setoid (BasedRay D) := raySetoid D
local instance visualSynchronousInverseRayDecidable : ∀ p : Prop, Decidable p :=
Classical.propDecidable
omit [TopologicalSpace G] [DiscreteTopology G] in
lemma inverse_ray_prefix (r : BasedRay D) {k l : ℕ} (hl : l ≤ k) :
D.graph.dist 1 ((r.val k)⁻¹*r.val (k-l)) = l ∧
D.graph.dist ((r.val k)⁻¹*r.val (k-l)) (r.val k)⁻¹ = k-l := by
have h₁ := D.dist_mul_left (r.val k) 1 ((r.val k)⁻¹*r.val (k-l))
have h₂ := D.dist_mul_left (r.val k) ((r.val k)⁻¹*r.val (k-l)) (r.val k)⁻¹
simp only [mul_one,mul_inv_cancel_left,mul_inv_cancel] at h₁ h₂
have hkk := r.property.2 k (k-l)
have hk₀ := r.property.2 (k-l) 0
rw [r.property.1,Nat.dist_zero_right] at hk₀
dsimp only [CayleyData.graph] at h₁ h₂ hkk hk₀ ⊢
constructor
· rw [← h₁,hkk,Nat.dist_eq_sub_of_le_right (Nat.sub_le _ _)]
omega
· rw [← h₂,hk₀]
theorem near_puncture_descendant_count {e H C : ℝ}
(he : 0 < e) (hH : 0 < H) (hC : 0 ≤ C) :
∃ N : ℕ, ∀ {ι : Type*} (S : Finset ι) (r : ι → BasedRay D)
(z a b : BasedRay D) (k l : ℕ), l ≤ k →
(∀ i ∈ S, e ≤ boundaryDist D hδ (Quotient.mk' (r i)) (Quotient.mk' z)) →
(∀ i ∈ S, ∀ j ∈ S, i ≠ j → H*visualScale δ k ≤
boundaryDist D hδ (Quotient.mk' (r i)) (Quotient.mk' (r j))) →
(∀ i ∈ S, boundaryDist D hδ (Quotient.mk' (r i)) (Quotient.mk' a) ≤
C*visualScale δ (k-l)) →
(∀ i ∈ S, boundaryDist D hδ
(boundaryTranslate D ⟨δ,hδ⟩ ((r i).val k)⁻¹ (Quotient.mk' z))
(Quotient.mk' b) ≤ visualScale δ l) →
S.card ≤ N := by
classical
obtain ⟨B,hB⟩ := inverse_puncture_prefix D hδ he
obtain ⟨B₁,hB₁⟩ := visual_synchronous_prefix D hδ 1 (by norm_num)
obtain ⟨B₂,hB₂⟩ := visual_synchronous_prefix D hδ C hC
obtain ⟨N₁,hN₁⟩ := uniform_graph_ball_card D (2*B+4*δ+2+B₁)
obtain ⟨N₂,hN₂⟩ := uniform_graph_ball_card D B₂
obtain ⟨Nv,hNv⟩ := vertex_assignment_multiplicity D hδ hH
refine ⟨Nv*(N₁*N₂),fun {ι} S r z a b k l hl hsep hcent hanc hpunc => ?_⟩
let u : ι → G := fun i => (r i).val (k-l)
let w : ι → G := fun i => ((r i).val k)⁻¹*u i
let p : ι → BasedRay D := fun i => translateRay D ⟨δ,hδ⟩ ((r i).val k)⁻¹ z
have hw (i : ι) (hi : i ∈ S) :
D.graph.dist (w i) (b.val l) ≤ 2*B+4*δ+2+B₁ := by
have hg : D.graph.dist 1 ((r i).val k)⁻¹ = k := by
have hh := D.dist_mul_left ((r i).val k) 1 ((r i).val k)⁻¹
have hk := (r i).property.2 k 0
rw [(r i).property.1,Nat.dist_zero_right] at hk
simpa only [mul_one,mul_inv_cancel,hk] using hh.symm
have hpre := inverse_ray_prefix D (r i) hl
have hnear := hB (r i) z k (hsep i hi)
have hh := prefix_of_near_ray D.connected hδ (p i) ((r i).val k)⁻¹ (w i)
k l B hl hg hpre.1 hpre.2 hnear
have hpi := hpunc i hi
have heq : boundaryTranslate D ⟨δ,hδ⟩ ((r i).val k)⁻¹ (Quotient.mk' z) =
Quotient.mk' (p i) := rfl
rw [heq] at hpi
have hsync := hB₁ (p i) b l (by simpa only [one_mul] using hpi)
exact (@SimpleGraph.Connected.dist_triangle _ D.graph (w i) ((p i).val l) (b.val l)
D.connected).trans (Nat.add_le_add hh hsync)
have hu (i : ι) (hi : i ∈ S) : D.graph.dist (u i) (a.val (k-l)) ≤ B₂ :=
hB₂ (r i) a (k-l) (hanc i hi)
have hwcard : (S.image w).card ≤ N₁ := hN₁ (b.val l) (S.image w) (by
intro y hy
obtain ⟨i,hi,rfl⟩ := mem_image.mp hy
rw [SimpleGraph.dist_comm]
exact hw i hi)
have hucard : (S.image u).card ≤ N₂ := hN₂ (a.val (k-l)) (S.image u) (by
intro y hy
obtain ⟨i,hi,rfl⟩ := mem_image.mp hy
rw [SimpleGraph.dist_comm]
exact hu i hi)
let f : ι → G × G := fun i => (w i,u i)
have hfcard : (S.image f).card ≤ N₁*N₂ := by
have hsub : S.image f ⊆ (S.image w) ×ˢ (S.image u) := by
intro y hy
obtain ⟨i,hi,rfl⟩ := mem_image.mp hy
exact mem_product.mpr ⟨mem_image.mpr ⟨i,hi,rfl⟩,mem_image.mpr ⟨i,hi,rfl⟩⟩
exact (Finset.card_le_card hsub).trans (by rw [card_product]; exact Nat.mul_le_mul hwcard hucard)
have hfib : ∀ y ∈ S.image f, (S.filter (fun i => f i = y)).card ≤ Nv := by
intro y _
apply (Finset.card_le_card (show S.filter (fun i => f i = y) ⊆
S.filter (fun i => (r i).val k = y.2*y.1⁻¹) from ?_)).trans (hNv S r k hcent _)
intro i hi
obtain ⟨hi,heq⟩ := mem_filter.mp hi
have hw := congrArg Prod.fst heq
have hu := congrArg Prod.snd heq
change w i = y.1 at hw
change u i = y.2 at hu
refine mem_filter.mpr ⟨hi,?_⟩
rw [← hw,← hu]
simp only [w,mul_inv_rev,inv_inv,mul_inv_cancel_left]
exact (card_le_mul_card_image S Nv hfib).trans (Nat.mul_le_mul_left Nv hfcard)
end CannonRelease
end
end
end
end
end
end
end OAI