Documentation

LeanPool.HighContrastHomogenization.Provider.Regularity.PrintOrderRoundedGenerationResponseScale

High-contrast homogenization: Provider.Regularity.PrintOrderRoundedGenerationResponseScale #

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

Selected-generation rounded response scale at the printed fractional order #

The boundary-layer discount and affine multiplier already used by the canonical response row apply uniformly to every admissible rounded generation.

theorem HCPolySupport.HighContrast.exists_summable_roundedGenerationResponseMaxRow_le_printOrderAffineTail_sq {d : ℕ} [NeZero d] {generation : ℤ} (hl : ↑(kZero d) ≤ generation) (a : CoeffSpace d) (abar : Mat d) (hS : Matrix.PosDef (symmPart abar)) (t : ℤ) (aRef : Book.Ch03.CoeffFamily d) (haRef : ∀ (R : TriadicCube d), (aRef.coeffOn R).toCoeffField = affineCoefficient (Selection.normalizedRoot (symmPart abar)) ⋯ ((normalizedCenteredCoeff a abar hS).coeffOn (Response.adaptedDomain ⋯ t)).toCoeffField) {g amplitude kappa x : ℝ} (hg : g ∈ Set.Ico 0 1) (hAmplitude : 0 ≤ amplitude) (hKappa : 0 < kappa) (hTail : ScalarIdentityPowerTail aRef (printCertificateOrder g) amplitude kappa x) (hx : 0 < x) (M : ℤ) (hActive : x ≤ 3 ^ M) :
∃ (G : ℕ) (Z : ℕ → Finset (Fin d → ℤ)), 3 ^ ↑G ≤ 1 + 3 * (100 / 99 * witnessEccentricity (symmPart abar) * √↑d) ∧ (∀ (l : ℕ), ↑(Z l) = {w : Fin d → ℤ | adaptedCellCenter (roundedGrid generation (symmPart abar)) (M - ↑l) w ∈ adaptedCell (roundedGrid generation (symmPart abar)) M}) ∧ (∀ (l : ℕ), (Z l).card = 3 ^ (d * l)) ∧ (Summable fun (l : ℕ) => Book.Ch02.geometricWeight (printCertificateOrder g) 2 l * Book.Ch02.finsetSupReal (Z l) fun (w : Fin d → ℤ) => roundedNormalizedDoubledResponseMaxAtGeneration generation a abar hS (M - ↑l) w) ∧ (∑' (l : ℕ), Book.Ch02.geometricWeight (printCertificateOrder g) 2 l * Book.Ch02.finsetSupReal (Z l) fun (w : Fin d → ℤ) => roundedNormalizedDoubledResponseMaxAtGeneration generation a abar hS (M - ↑l) w) ≤ (printOrderRoundedResponseAffineConstant d g * witnessEccentricity (symmPart abar) * (amplitude * (3 ^ M / x) ^ (-kappa))) ^ 2

The outer response row is controlled by the printed-order affine loss.

theorem HCPolySupport.HighContrast.exists_summable_roundedGenerationResponseMaxRow_le_printOrderEffectiveScale_powerTail_sq {d : ℕ} [NeZero d] {generation : ℤ} (hl : ↑(kZero d) ≤ generation) (a : CoeffSpace d) (abar : Mat d) (hS : Matrix.PosDef (symmPart abar)) (t : ℤ) (aRef : Book.Ch03.CoeffFamily d) (haRef : ∀ (R : TriadicCube d), (aRef.coeffOn R).toCoeffField = affineCoefficient (Selection.normalizedRoot (symmPart abar)) ⋯ ((normalizedCenteredCoeff a abar hS).coeffOn (Response.adaptedDomain ⋯ t)).toCoeffField) {g amplitude kappa x : ℝ} (hg : g ∈ Set.Ico 0 1) (hAmplitude : 0 ≤ amplitude) (hKappa : 0 < kappa) (hTail : ScalarIdentityPowerTail aRef (printCertificateOrder g) amplitude kappa x) (hx : 0 < x) (M : ℤ) (hActive : roundedAffineEffectiveScale (printOrderRoundedResponseAffineConstant d g) (specBound (symmPart abar) * specBound (symmPart abar)⁻¹) kappa x ≤ 3 ^ M) :
∃ (G : ℕ) (Z : ℕ → Finset (Fin d → ℤ)), 3 ^ ↑G ≤ 1 + 3 * (100 / 99 * witnessEccentricity (symmPart abar) * √↑d) ∧ (∀ (l : ℕ), ↑(Z l) = {w : Fin d → ℤ | adaptedCellCenter (roundedGrid generation (symmPart abar)) (M - ↑l) w ∈ adaptedCell (roundedGrid generation (symmPart abar)) M}) ∧ (∀ (l : ℕ), (Z l).card = 3 ^ (d * l)) ∧ (Summable fun (l : ℕ) => Book.Ch02.geometricWeight (printCertificateOrder g) 2 l * Book.Ch02.finsetSupReal (Z l) fun (w : Fin d → ℤ) => roundedNormalizedDoubledResponseMaxAtGeneration generation a abar hS (M - ↑l) w) ∧ (∑' (l : ℕ), Book.Ch02.geometricWeight (printCertificateOrder g) 2 l * Book.Ch02.finsetSupReal (Z l) fun (w : Fin d → ℤ) => roundedNormalizedDoubledResponseMaxAtGeneration generation a abar hS (M - ↑l) w) ≤ (amplitude * (3 ^ M / roundedAffineEffectiveScale (printOrderRoundedResponseAffineConstant d g) (specBound (symmPart abar) * specBound (symmPart abar)⁻¹) kappa x) ^ (-kappa)) ^ 2

The printed-order affine scale absorbs the rounded response loss.