openAI-math-dataset / lean /OAI /Combinatorics /MatrixRemoval /AnchorPatternBridge.lean
roseking's picture
Add lean source and binary files (part 47)
e5a97c6 verified
Raw History Blame Contribute Delete
6 kB
import Mathlib
namespace OAI
/-!
# Structural bridges for anchor pinning
Group uniqueness and no-shattering give pinning criteria independent of
the host construction.
-/
universe uI uJ uR uC uG
namespace Problem348.AnchorBridge
/-- Distinct pattern rows prevent a copy from repeating any host row class
whose members have identical entries. -/
theorem row_labels_injective {I : Type uI} {J : Type uJ} {R : Type uR} {C : Type uC} {G : Type uG}
(H : I → J → Bool) (A : R → C → Bool)
(row : I → R) (col : J → C) (label : R → G)
(hcopy : ∀ i j, A (row i) (col j) = H i j)
(hrows : Function.Injective H)
(hlabel : ∀ r s, label r = label s → A r = A s) :
Function.Injective (fun i => label (row i)) := by
intro i j hij
apply hrows
funext c
rw [← hcopy, ← hcopy]
exact congrFun (hlabel (row i) (row j) hij) (col c)
/-- The column version of `row_labels_injective`. -/
theorem col_labels_injective {I : Type uI} {J : Type uJ} {R : Type uR} {C : Type uC} {G : Type uG}
(H : I → J → Bool) (A : R → C → Bool)
(row : I → R) (col : J → C) (label : C → G)
(hcopy : ∀ i j, A (row i) (col j) = H i j)
(hcols : Function.Injective (fun j => fun i => H i j))
(hlabel : ∀ c d, label c = label d → ∀ r, A r c = A r d) :
Function.Injective (fun j => label (col j)) := by
intro i j hij
apply hcols
funext r
change H r i = H r j
rw [← hcopy, ← hcopy]
exact hlabel (col i) (col j) hij (row r)
/-- The 32 rows carrying all binary words in the fixed anchor. -/
def wordRow (a : Fin 32) : Fin 64 := ⟨32 + a.val, by omega⟩
/-- The five columns carrying the word coordinates in the fixed anchor. -/
def wordCol (b : Fin 5) : Fin 64 := ⟨59 + b.val, by omega⟩
@[simp] theorem wordCol_injective : Function.Injective wordCol := by
intro a b hab
apply Fin.ext
have := congrArg Fin.val hab
simp only [wordCol] at this
omega
/-- If both axes had at most 32 initial anchor positions, the complete-word
submatrix would shatter five non-anchor columns. Thus the position at index
32 is an anchor on at least one axis. -/
theorem anchor_at_32_or_anchor_at_32 {R : Type uR} {C : Type uC}
(H : Fin 64 → Fin 64 → Bool) (A : R → C → Bool)
(row : Fin 64 → R) (col : Fin 64 → C)
(anchorR : R → Prop) (anchorC : C → Prop)
(hcopy : ∀ i j, A (row i) (col j) = H i j)
(hcol : Function.Injective col)
(hprefixR : ∀ i j, i ≤ j → anchorR (row j) → anchorR (row i))
(hprefixC : ∀ i j, i ≤ j → anchorC (col j) → anchorC (col i))
(hwords : ∀ w : Fin 5 → Bool, ∃ a : Fin 32,
∀ b : Fin 5, H (wordRow a) (wordCol b) = w b)
(hnoshatter : ∀ cs : Fin 5 → C, Function.Injective cs →
(∀ b, ¬ anchorC (cs b)) → ∃ w : Fin 5 → Bool,
∀ r, ¬ anchorR r → ∃ b, A r (cs b) ≠ w b) :
anchorR (row ⟨32, by decide⟩) ∨ anchorC (col ⟨32, by decide⟩) := by
classical
by_contra h
push Not at h
have hr : ∀ a : Fin 32, ¬ anchorR (row (wordRow a)) := by
intro a ha
apply h.1
exact hprefixR _ _ (by simp only [wordRow, Fin.le_iff_val_le_val]; omega) ha
have hc : ∀ b : Fin 5, ¬ anchorC (col (wordCol b)) := by
intro b hb
apply h.2
exact hprefixC _ _ (by simp only [wordCol, Fin.le_iff_val_le_val]; omega) hb
obtain ⟨w, hw⟩ := hnoshatter (fun b => col (wordCol b))
(hcol.comp wordCol_injective) hc
obtain ⟨a, ha⟩ := hwords w
obtain ⟨b, hb⟩ := hw (row (wordRow a)) (hr a)
exact hb ((hcopy _ _).trans (ha b))
/-- A downward-closed predicate holding at index `m` holds at at least
`m+1` indices. -/
theorem prefix_card_lower_bound {n : ℕ} (P : Fin n → Prop) [DecidablePred P]
(m : Fin n) (hprefix : ∀ i j, i ≤ j → P j → P i) (hm : P m) :
m.val + 1 ≤ (Finset.univ.filter P).card := by
classical
calc
m.val + 1 = (Finset.Iic m).card := by simp
_ ≤ (Finset.univ.filter P).card := by
apply Finset.card_le_card
intro i hi
exact Finset.mem_filter.mpr ⟨Finset.mem_univ i,
hprefix i m (Finset.mem_Iic.mp hi) hm⟩
/-- An increasing self-map of a finite ordered interval uses every label
in the prescribed order. -/
theorem increasing_labels_eq_id {n : ℕ} (f : Fin n → Fin n)
(hf : StrictMono f) : f = id := by
funext i
exact le_antisymm (hf.le_id i) (hf.id_le i)
/-- Any class of host rows with identical entries is used at most once
in a copy with distinct pattern rows. In particular this applies to the
dummy row class. -/
theorem card_selected_row_class_le_one {I : Type uI} {J : Type uJ} {R : Type uR} {C : Type uC}
[Fintype I] (H : I → J → Bool) (A : R → C → Bool)
(row : I → R) (col : J → C) (P : R → Prop) [DecidablePred P]
(hcopy : ∀ i j, A (row i) (col j) = H i j)
(hrows : Function.Injective H)
(hclass : ∀ r s, P r → P s → A r = A s) :
(Finset.univ.filter (fun i => P (row i))).card ≤ 1 := by
classical
apply Finset.card_le_one.mpr
intro i hi j hj
apply hrows
funext c
rw [← hcopy, ← hcopy]
exact congrFun (hclass (row i) (row j)
(Finset.mem_filter.mp hi).2 (Finset.mem_filter.mp hj).2) (col c)
/-- Column version of the single-use class bound. -/
theorem card_selected_col_class_le_one {I : Type uI} {J : Type uJ} {R : Type uR} {C : Type uC}
[Fintype J] (H : I → J → Bool) (A : R → C → Bool)
(row : I → R) (col : J → C) (P : C → Prop) [DecidablePred P]
(hcopy : ∀ i j, A (row i) (col j) = H i j)
(hcols : Function.Injective (fun j => fun i => H i j))
(hclass : ∀ c d, P c → P d → ∀ r, A r c = A r d) :
(Finset.univ.filter (fun j => P (col j))).card ≤ 1 := by
classical
apply Finset.card_le_one.mpr
intro i hi j hj
apply hcols
funext r
change H r i = H r j
rw [← hcopy, ← hcopy]
exact hclass (col i) (col j)
(Finset.mem_filter.mp hi).2 (Finset.mem_filter.mp hj).2 (row r)
end Problem348.AnchorBridge
end OAI