import OAI.Combinatorics.GraphThreshold.GraphSupport import OAI.Combinatorics.GraphThreshold.GraphProbability import OAI.Combinatorics.GraphThreshold.CopyOrbits import OAI.Combinatorics.GraphThreshold.GraphExpectedCopies namespace OAI noncomputable section open scoped Pointwise namespace LeanBlast.SecondKahnKalai variable {n : ℕ} theorem mem_copyOrbit_iff_edgeSupportGraph_iso {S T : Finset (Edge n)} : T ∈ copyOrbit (Equiv.Perm (Fin n)) S ↔ Nonempty (edgeSupportGraph S ≃g edgeSupportGraph T) := by rw [mem_copyOrbit_iff] constructor · rintro ⟨σ, rfl⟩ exact ⟨edgeSupportGraph_smulIso σ S⟩ · rintro ⟨e⟩ exact exists_smul_eq_of_edgeSupportGraph_iso e /-- Distinct edge orbits correspond bijectively to actual unlabelled subgraphs. The target has no isolated vertices, so its image subgraph is uniquely determined by its edges. This also covers the empty graph, with its empty vertex set. -/ def orbitSubgraphEquiv (S : Finset (Edge n)) : OrbitCopy (Equiv.Perm (Fin n)) S ≃ {K : (⊤ : SimpleGraph (Fin n)).Subgraph // Nonempty (edgeSupportGraph S ≃g K.coe)} where toFun T := ⟨edgeSubgraph T.val, mem_copyOrbit_iff_edgeSupportGraph_iso.mp T.property⟩ invFun K := ⟨graphEdges K.val.spanningCoe, by apply mem_copyOrbit_iff_edgeSupportGraph_iso.mpr obtain ⟨e⟩ := K.property have hK := subgraph_eq_edgeSubgraph_of_iso K.val e change Nonempty (edgeSupportGraph S ≃g (edgeSubgraph _).coe) rw [← hK] exact ⟨e⟩⟩ left_inv T := by apply Subtype.ext exact graphEdges_graphOfEdges T.val right_inv K := by apply Subtype.ext exact (subgraph_eq_edgeSubgraph_of_iso K.val K.property.some).symm theorem orbitCount_eq_copyCount (S : Finset (Edge n)) : orbitCount (Equiv.Perm (Fin n)) S = (⊤ : SimpleGraph (Fin n)).copyCount (edgeSupportGraph S) := by classical calc orbitCount (Equiv.Perm (Fin n)) S = Fintype.card (OrbitCopy (Equiv.Perm (Fin n)) S) := (card_OrbitCopy _ _).symm _ = Fintype.card {K : (⊤ : SimpleGraph (Fin n)).Subgraph // Nonempty (edgeSupportGraph S ≃g K.coe)} := Fintype.card_congr (orbitSubgraphEquiv S) _ = _ := by rw [SimpleGraph.copyCount, Fintype.card_subtype] apply congrArg Finset.card ext K simp only [Finset.mem_filter, Finset.mem_univ, true_and] theorem expectedCopies_edgeSupportGraph (p : ℝ) (S : Finset (Edge n)) : expectedCopies n p (edgeSupportGraph S) = (orbitCount (Equiv.Perm (Fin n)) S : ℝ) * p ^ S.card := by rw [expectedCopies_formula, edgeCount_edgeSupportGraph, orbitCount_eq_copyCount] /-- The internal orbit constraint follows from an actual subgraph constraint of the public expectation threshold, with no multiplicity factor. -/ theorem orbit_expectation_constraint {V : Type*} (f : V ↪ Fin n) (H : SimpleGraph V) (q : ℝ) (hq : ∀ F : H.Subgraph, (1 / 2 : ℝ) ≤ expectedCopies n q F.coe) (S : Finset (Edge n)) (hS : S ⊆ graphEdges (H.map f)) : (1 / 2 : ℝ) ≤ (orbitCount (Equiv.Perm (Fin n)) S : ℝ) * q ^ S.card := by obtain ⟨F, ⟨e⟩⟩ := exists_subgraph_iso_edgeSupportGraph f H S hS rw [← expectedCopies_edgeSupportGraph, expectedCopies_iso n q e] exact hq F theorem copyCount_pos_iff_orbit_subset {V : Type*} (f : V ↪ Fin n) (H : SimpleGraph V) (W : Finset (Edge n)) : 0 < (graphOfEdges W).copyCount H ↔ ∃ E ∈ copyOrbit (Equiv.Perm (Fin n)) (graphEdges (H.map f)), E ⊆ W := by rw [copyCount_pos_iff_exists_smul_subset f H W] constructor · rintro ⟨σ, hσ⟩ exact ⟨_, (mem_copyOrbit_iff _).mpr ⟨σ, rfl⟩, hσ⟩ · rintro ⟨E, hE, hEW⟩ obtain ⟨σ, rfl⟩ := (mem_copyOrbit_iff _).mp hE exact ⟨σ, hEW⟩ theorem exists_graphRepresentationWithConstraints {V : Type*} [Fintype V] (n : ℕ) (H : SimpleGraph V) (hcard : Fintype.card V ≤ n) (q : ℝ) (hq : ∀ F : H.Subgraph, (1 / 2 : ℝ) ≤ expectedCopies n q F.coe) : ∃ E : Finset (Edge n), E.card = edgeCount H ∧ (∀ W : Finset (Edge n), (0 < (graphOfEdges W).copyCount H ↔ ∃ T ∈ copyOrbit (Equiv.Perm (Fin n)) E, T ⊆ W)) ∧ (∀ S : Finset (Edge n), S ⊆ E → (1 / 2 : ℝ) ≤ (orbitCount (Equiv.Perm (Fin n)) S : ℝ) * q ^ S.card) := by obtain ⟨f⟩ := Function.Embedding.nonempty_of_card_le (show Fintype.card V ≤ Fintype.card (Fin n) by simpa using hcard) exact ⟨graphEdges (H.map f), card_mappedGraphEdges f H, copyCount_pos_iff_orbit_subset f H, orbit_expectation_constraint f H q hq⟩ end LeanBlast.SecondKahnKalai end end OAI