import OAI.Combinatorics.CycleDecomposition.BatchIndexing namespace OAI section open Filter Asymptotics Real open scoped Topology noncomputable section open MeasureTheory ProbabilityTheory Finset section namespace ErdosGallai noncomputable section open Finset theorem prefix_selection (S : ℕ → ℝ) (hS : ∀ j, 0 ≤ S j) (K N : ℕ) (hsupp : ∀ j, N ≤ j → S j = 0) (hne : ∃ j, 0 < S j) : ∃ i, 0 < S i ∧ ∑ j ∈ range (i+K+1), S j ≤ 4*(K+1:ℝ)*S i := by let A := (range N).filter (fun j => 0 < S j) have hA : A.Nonempty := by obtain ⟨j,hj⟩ := hne exact ⟨j, by simp only [A, mem_filter, mem_range] exact ⟨by by_contra hn; rw [hsupp j (by omega)] at hj; linarith, hj⟩⟩ let i₀ := A.min' hA have hi₀A : i₀ ∈ A := A.min'_mem hA have hi₀pos : 0 < S i₀ := (mem_filter.mp hi₀A).2 have hstart : (∑ j ∈ range (i₀+1), S j) = S i₀ := by apply sum_eq_single i₀ · intro j hj hji by_contra hjzero have hjpos : 0 < S j := lt_of_le_of_ne (hS j) (Ne.symm hjzero) have hjA : j ∈ A := by simp only [A, mem_filter, mem_range] refine ⟨?_, hjpos⟩ have hiN := (mem_range.mp (mem_filter.mp hi₀A).1) have hjle := mem_range.mp hj omega have hmin : i₀ ≤ j := A.min'_le j hjA have hjle := mem_range.mp hj omega · intro hi exact (hi (by simp)).elim let B := A.filter (fun i => (∑ j ∈ range (i+1), S j) ≤ 2*(K+1:ℝ)*S i) have hB : B.Nonempty := by refine ⟨i₀, mem_filter.mpr ⟨hi₀A, ?_⟩⟩ rw [hstart] have hK : (0:ℝ) ≤ K := Nat.cast_nonneg K nlinarith let i := B.max' hB have hiB : i ∈ B := B.max'_mem hB have hiA : i ∈ A := (mem_filter.mp hiB).1 have hipos : 0 < S i := (mem_filter.mp hiA).2 have hinv : (∑ j ∈ range (i+1), S j) ≤ 2*(K+1:ℝ)*S i := (mem_filter.mp hiB).2 refine ⟨i, hipos, ?_⟩ by_contra hfail have hfail' : 4*(K+1:ℝ)*S i < ∑ j ∈ range (i+K+1), S j := lt_of_not_ge hfail have hwindow : 2*(K+1:ℝ)*S i < ∑ j ∈ Ico (i+1) (i+K+1), S j := by have hsplit := Finset.sum_range_add_sum_Ico S (by omega : i+1 ≤ i+K+1) linarith have hwindow_nonempty : (Ico (i+1) (i+K+1)).Nonempty := by by_contra hh rw [Finset.not_nonempty_iff_eq_empty.mp hh] at hwindow simp only [sum_empty] at hwindow have hK : (0:ℝ) ≤ K := Nat.cast_nonneg K nlinarith obtain ⟨j,hjw,hmax⟩ := Finset.exists_max_image (Ico (i+1) (i+K+1)) S hwindow_nonempty have hsum : (∑ l ∈ Ico (i+1) (i+K+1), S l) ≤ (K:ℝ)*S j := by calc _ ≤ ∑ _l ∈ Ico (i+1) (i+K+1), S j := sum_le_sum hmax _ = _ := by simp [] have hdouble : 2*S i < S j := by have hK : (0:ℝ) ≤ K := Nat.cast_nonneg K nlinarith have hjpos : 0 < S j := by linarith have hjA : j ∈ A := by simp only [A, mem_filter, mem_range] exact ⟨by by_contra hjN; rw [hsupp j (by omega)] at hjpos; linarith, hjpos⟩ have hjbounds := Finset.mem_Ico.mp hjw have hpartial : (∑ l ∈ Ico (i+1) (j+1), S l) ≤ (K:ℝ)*S j := by calc _ ≤ ∑ l ∈ Ico (i+1) (i+K+1), S l := by apply sum_le_sum_of_subset_of_nonneg · intro l hl simp only [mem_Ico] at hl ⊢ omega · intro l _ _; exact hS l _ ≤ _ := hsum have hjinv : (∑ l ∈ range (j+1), S l) ≤ 2*(K+1:ℝ)*S j := by have hsplit := Finset.sum_range_add_sum_Ico S (by omega : i+1 ≤ j+1) have hK : (0:ℝ) ≤ K := Nat.cast_nonneg K nlinarith have hjB : j ∈ B := mem_filter.mpr ⟨hjA,hjinv⟩ have hji : j ≤ i := B.le_max' j hjB omega end end ErdosGallai end namespace ErdosGallai def MainStatement : Prop := ∃ C : ℝ, 0 < C ∧ ∀ (n : ℕ) (G : SimpleGraph (Fin n)), ∃ k : ℕ, EdgeDecomposition G k ∧ (k : ℝ) ≤ C * n end ErdosGallai end end end OAI