Documentation

ConNF.Model.InductionStatement

New file #

In this file...

Main declarations #

structure ConNF.Motive [Params] (α : Λ) :
Type (u + 1)
Instances For
    Equations
    Instances For
      theorem ConNF.pos_tangle_bot [Params] {D : ModelData ⊥} (t : Tangle ⊥) (a : Atom) (A : ⊥ ↝ ⊥) (ha : a ∈ (t.support ⇘. A)ᴬ) :
      pos a < pos t
      structure ConNF.Hypothesis [Params] {α : Λ} (M : Motive α) (N : (β : Λ) → ↑β < ↑α → Motive β) :
      Instances For
        def ConNF.castTSet [Params] {α β : TypeIndex} {D₁ : ModelData α} {D₂ : ModelData β} (h₁ : α = β) (h₂ : D₁ ≍ D₂) :
        TSet α ≃ TSet β
        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          def ConNF.castAllPerm [Params] {α β : TypeIndex} {D₁ : ModelData α} {D₂ : ModelData β} (h₁ : α = β) (h₂ : D₁ ≍ D₂) :
          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            def ConNF.castTangle [Params] {α β : TypeIndex} {D₁ : ModelData α} {D₂ : ModelData β} (h₁ : α = β) (h₂ : D₁ ≍ D₂) :
            Equations
            • One or more equations did not get rendered due to their size.
            Instances For
              theorem ConNF.castTSet_forget [Params] {α : TypeIndex} {D₁ D₂ : ModelData α} (h₂ : D₁ ≍ D₂) (ρ : TSet α) :
              theorem ConNF.castAllPerm_forget [Params] {α : TypeIndex} {D₁ D₂ : ModelData α} (h₂ : D₁ ≍ D₂) (ρ : AllPerm α) :
              theorem ConNF.castTangle_support [Params] {α : TypeIndex} {D₁ D₂ : ModelData α} (h₂ : D₁ ≍ D₂) (t : Tangle α) :
              ((castTangle ⋯ h₂) t).support = t.support
              theorem ConNF.castAllPerm_smul [Params] {α : TypeIndex} {D₁ D₂ : ModelData α} (h₂ : D₁ ≍ D₂) (ρ : AllPerm α) (t : Tangle α) :
              ρ • (castTangle ⋯ h₂) t = (castTangle ⋯ h₂) ((castAllPerm ⋯ h₂).symm ρ • t)
              theorem ConNF.castTangle_pos [Params] {α β : TypeIndex} {D₁ : ModelData α} {D₂ : ModelData β} {E₁ : Position (Tangle α)} {E₂ : Position (Tangle β)} (h₁ : α = β) (h₂ : D₁ ≍ D₂) (h₃ : E₁ ≍ E₂) (t : Tangle α) :
              pos ((castTangle h₁ h₂) t) = pos t
              theorem ConNF.castTangle_fuzz [Params] {α : TypeIndex} {D₁ D₂ : ModelData α} {E₁ : Position (Tangle α)} {E₂ : Position (Tangle α)} (h₂ : D₁ ≍ D₂) (h₃ : E₁ ≍ E₂) (t : Tangle α) {β : Λ} (hαβ : α ≠ ↑β) :
              fuzz hαβ ((castTangle ⋯ h₂) t) = fuzz hαβ t