Skip to content

Formalize Karlin's weak sign-regular variation-diminishing theorem #293

Description

@PerAlexandersson

Source

Karlin, Total Positivity, Vol. I, Chapter V, §1, Theorem 1.4 (printed p. 223).

For a sign-regular matrix U of rank r, the forward theorem is

S⁻(Uc) ≤ min(r - 1, S⁻(c)).

The source says this follows from Theorem 1.3 exactly as the strict sign-block theorem V.1.2 follows from V.1.1. The converse under full column rank belongs to a later dependency and is not required for the initial ASW forward route.

Required source proof route

  1. Depend on the completed V.1.3 rank-sensitive sign-consistent theorem; do not assume the rank bound separately.
  2. Split c into S⁻(c)+1 consecutive nonempty same-sign blocks using Fin.SignBlockDecomposition.
  3. Aggregate columns with weights |c_j|, exactly as in V.1.2, and express Uc as the aggregate matrix applied to the cast block-sign vector.
  4. Use sign-regularity of U, determinant multilinearity, and the existing fiberwise determinant expansion to prove that the aggregate matrix is sign-consistent of the relevant order. In the weak case summands may vanish; no positive-selector hypothesis is available or required.
  5. Apply V.1.3 to the aggregate matrix to obtain the S⁻(c) bound. Apply V.1.3 directly to U for the rank U - 1 bound, then combine them with min.
  6. Translate the project convention S⁻(0)=0 explicitly.

Proposed main declaration

theorem Matrix.IsSignRegular.signVariations_mulVec_le
    {m n r : ℕ} {U : Matrix (Fin n) (Fin m) ℝ}
    (hU : U.IsSignRegular)
    (hrank : Module.finrank ℝ (LinearMap.range U.mulVecLin) = r)
    (c : Fin m → ℝ) :
    Fin.signVariations (U.mulVec c) ≤
      min (r - 1) (Fin.signVariations c) := by
  ...

IsSignRegular should be a genuine reusable property definition quantifying the common weak sign of minors of every order; it must not be a theorem placeholder.

Acceptance criteria

  • Follows the paper’s V.1.3 → V.1.4 sign-block route.
  • Reuses Matrix.det_submatrix_fiberwise_sum and Fin.SignBlockDecomposition.
  • No sorry, axioms, or theorem-shaped def ...Statement : Prop.
  • Proves both components of the min bound.
  • Strong enough to replace the ASW forward variation-diminishing assumption after the Toeplitz matrix is shown TNN/sign-regular.
  • Focused and full CI are green.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions