Documentation

LeanPool.HighContrastHomogenization.Entry.Geometry.AdaptedCell

High-contrast homogenization: Entry.Geometry.AdaptedCell #

Imported from the Apache-2.0 HighContrastHomogenization development at commit 7a13dbcd8d6609264a713373f5c69ceeac870472.

Adapted cells #

The adapted cell ⋄_j^q = q □_j and its translate W = y + ⋄_j^q of the paper, together with the two facts about W that l.source.whitney uses: its volume under the change of variables x = y + q v, and the strip estimate |{x ∈ W : dist(x, ∂W) ≤ s}| / |W| ≤ 2 d |q⁻¹| s 3^{-j}.

The definitions adaptedCell and adaptedCellTranslate live in HCPoly.Setup.Geometry (imported here), so that they can be used without these proofs. The bound |q⁻¹| ≤ K of e.rounded.grid.bounds enters only through the hypothesis InverseNormLE q K, i.e. |v| ≤ K |q v| for all v (stated with squared Euclidean norms).

Definitions #

The operator-norm bound |q⁻¹| ≤ K, in the form |v|² ≤ K² |q v|² for every v.

Equations
Instances For

    Volume under the change of variables #

    Invertibility from the inverse bound #

    theorem HCPolySupport.HighContrast.Geometry.adaptedCellTranslate_eq_preimage {d : ℕ} {q : Mat d} (hq : IsUnit q) (j : ℤ) (y : Vec d) :
    adaptedCellTranslate q j y = (fun (x : Fin d → ℝ) => q⁻¹.mulVec (x - y)) ⁻¹' centeredCube d j

    For invertible q, the translate y + q □_j is the preimage of □_j under x ↦ q⁻¹ (x - y).

    The strip estimate #

    The part of □_j within t of its boundary, coordinatewise: points of □_j with some coordinate of absolute value at least 3^j/2 - t.

    Equations
    Instances For

      One coordinate strip of □_j of width t, as a product set.

      Equations
      • One or more equations did not get rendered due to their size.
      Instances For
        theorem HCPolySupport.HighContrast.Geometry.mem_image_boundaryStrips {d : ℕ} {q : Mat d} {K : ℝ} (hK : 0 ≤ K) (hq : InverseNormLE q K) {j : ℤ} {y x x' : Vec d} {s : ℝ} (hs : 0 ≤ s) (hx : x ∈ adaptedCellTranslate q j y) (hx' : x' ∉ adaptedCellTranslate q j y) (hdist : ∑ i : Fin d, (x i - x' i) ^ 2 ≤ s ^ 2) :
        x ∈ (fun (v : Vec d) => y + matVecMul q v) '' boundaryStrips d j (K * s)

        Points of W within Euclidean distance s of a point outside W lie in the image of the boundary strips of width K s.

        theorem HCPolySupport.HighContrast.Geometry.volume_le_of_near_complement {d : ℕ} [NeZero d] {q : Mat d} {K : ℝ} (hK : 0 ≤ K) (hq : InverseNormLE q K) {j : ℤ} {y : Vec d} {A : Set (Vec d)} {s : ℝ} (hs : 0 ≤ s) (hAW : A ⊆ adaptedCellTranslate q j y) (hA : ∀ x ∈ A, ∃ x' ∉ adaptedCellTranslate q j y, ∑ i : Fin d, (x i - x' i) ^ 2 ≤ s ^ 2) :

        The strip estimate. If every point of A ⊆ W is within Euclidean distance s of a point outside W, then |A| ≤ 2 d K s 3^{-j} |W|, where |q⁻¹| ≤ K.