Documentation

LeanPool.MatrixMultiplicationAllFields.LinearAlgebra.MatrixMultiplication.AuxiliarySeparation.Spectrum.StateObstruction

+# Finite obstructions to normalized additive states

The state equations are homogeneous integer inequalities, except for the normalization at the unit. Their finite certificates are interpreted in the additive group completion, preserving catalysts in the original semiring.

The homogeneous inequalities defining a normalized additive state. Both signs of the additive relation are included.

Instances For

    Integer coefficient vector of a homogeneous state constraint.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      Evaluation of a finite integer coefficient vector on a real-valued function.

      Equations
      Instances For
        theorem MatrixMultiplicationAllFields.MatrixMultiplication.AuxiliarySeparation.normalizedStates_iff_stateConstraint {S : Type u_1} [CommSemiring S] [Preorder S] (R : S → ℕ) (d : S) (k : ℕ) (f : S → ℝ) :
        f ∈ normalizedStates (fun (x : S) => ↑(R x)) d ↑k ↔ f 1 = 1 ∧ ∀ (c : StateConstraint S), 0 ≤ (statePairing f) (stateConstraintVector R d k c)

        The semiring element providing the upper endpoint of a state constraint.

        Equations
        Instances For

          The multiplier of the distinguished detector difference.

          Equations
          Instances For
            theorem MatrixMultiplicationAllFields.MatrixMultiplication.AuxiliarySeparation.no_stateConstraint_certificate {S : Type u_1} [CommSemiring S] [Preorder S] (R : S → ℕ) (d : S) (k : ℕ) (hzero : ∀ (x : S), 0 ≤ x) (hbound : ∀ (x : S), x ≤ ↑(R x)) (hadd : ∀ (a b c e : S), a ≤ b → c ≤ e → a + c ≤ b + e) (hno : ∀ (D s : S) (m : ℕ), 0 < m → ¬D + ↑m + d * s ≤ D + ↑k * s) {I : Type u_2} [Fintype I] (cs : I → StateConstraint S) :
            ¬∃ (m : ℕ), 0 < m ∧ ∃ (n : I → ℕ), ∑ i : I, n i • stateConstraintVector R d k (cs i) = -↑m • Finsupp.single 1 1

            In the absence of a catalytic obstruction, no positive multiple of the negative unit is a natural combination of state-constraint vectors.

            theorem MatrixMultiplicationAllFields.MatrixMultiplication.AuxiliarySeparation.finite_stateConstraint_solution {S : Type u_1} [CommSemiring S] [Preorder S] (R : S → ℕ) (d : S) (k : ℕ) (hzero : ∀ (x : S), 0 ≤ x) (hbound : ∀ (x : S), x ≤ ↑(R x)) (hadd : ∀ (a b c e : S), a ≤ b → c ≤ e → a + c ≤ b + e) (hno : ∀ (D s : S) (m : ℕ), 0 < m → ¬D + ↑m + d * s ≤ D + ↑k * s) {I : Type u_2} [Finite I] (cs : I → StateConstraint S) (H : Finset S) (hOne : 1 ∈ H) (hH : ∀ (i : I), (stateConstraintVector R d k (cs i)).support ⊆ H) :
            ∃ (f : S → ℝ), f 1 = 1 ∧ (∀ (i : I), 0 ≤ (statePairing f) (stateConstraintVector R d k (cs i))) ∧ ∀ x ∉ H, f x = 0

            Every finite family of constraints supported on H has a normalized solution that vanishes outside H, provided no positive catalyst exists.

            theorem MatrixMultiplicationAllFields.MatrixMultiplication.AuxiliarySeparation.bounded_finite_stateConstraint_solution {S : Type u_1} [CommSemiring S] [Preorder S] (R : S → ℕ) (d : S) (k : ℕ) (hzero : ∀ (x : S), 0 ≤ x) (hbound : ∀ (x : S), x ≤ ↑(R x)) (hadd : ∀ (a b c e : S), a ≤ b → c ≤ e → a + c ≤ b + e) (hno : ∀ (D s : S) (m : ℕ), 0 < m → ¬D + ↑m + d * s ≤ D + ↑k * s) (F : Finset (StateConstraint S)) :
            ∃ (f : S → ℝ), (∀ (x : S), 0 ≤ f x ∧ f x ≤ ↑(R x)) ∧ f 1 = 1 ∧ ∀ c ∈ F, 0 ≤ (statePairing f) (stateConstraintVector R d k c)

            A finite collection of state constraints can be solved while keeping all coordinates in their rank intervals.

            theorem MatrixMultiplicationAllFields.MatrixMultiplication.AuxiliarySeparation.normalizedStates_nonempty_of_no_catalyst {S : Type u_1} [CommSemiring S] [Preorder S] (R : S → ℕ) (d : S) (k : ℕ) (hzero : ∀ (x : S), 0 ≤ x) (hbound : ∀ (x : S), x ≤ ↑(R x)) (hadd : ∀ (a b c e : S), a ≤ b → c ≤ e → a + c ≤ b + e) (hno : ∀ (D s : S) (m : ℕ), 0 < m → ¬D + ↑m + d * s ≤ D + ↑k * s) :
            (normalizedStates (fun (x : S) => ↑(R x)) d ↑k).Nonempty

            Absence of positive scalar-gain catalysts guarantees the existence of a normalized additive monotone state satisfying the detector inequality.