Documentation

LeanPool.HighContrastHomogenization.Entry.Analysis.PositiveGap

High-contrast homogenization: Entry.Analysis.PositiveGap #

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

PositiveGap random-field assembly #

This module assembles the ordered random difference and the deterministic mean-gap receipts that the positive-gap estimates consume, and proves the nonlinear, Young absorption and centered forms of the positive-gap estimate l.fixed.geometry.positive.gap.

theorem HCPolySupport.HighContrast.Analysis.positiveGap_ordered_data {d : ℕ} {P : MeasureTheory.Measure (CoeffSpace d)} [MeasureTheory.IsFiniteMeasure P] {N : ℝ} (hN : 1 ≤ N) {F G : CoeffSpace d → BlockMat d} (hF : SchattenMemLp P N F) (hG : SchattenMemLp P N G) (hFpos : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (ofFullBlockMat 0) (F a)) (hFG : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (F a) (G a)) :
have D := fun (a : CoeffSpace d) => blockSub (G a) (F a); have MF := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (F a) α β ∂P); have MG := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (G a) α β ∂P); SchattenMemLp P N D ∧ (∀ᵐ (a : CoeffSpace d) ∂P, Matrix.PosSemidef (toFullBlockMat (D a)) ∧ BlockMatLoewnerLE (D a) (G a)) ∧ ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (D a) α β ∂P) = blockSub MG MF ∧ Matrix.PosSemidef (toFullBlockMat MF) ∧ Matrix.PosSemidef (toFullBlockMat MG) ∧ Matrix.PosSemidef (toFullBlockMat (blockSub MG MF)) ∧ BlockMatLoewnerLE (blockSub MG MF) MG

The ordered difference D = G - F and its entrywise mean receipts.

theorem HCPolySupport.HighContrast.Analysis.positiveGap_mean_gap_bounds {d : ℕ} {P : MeasureTheory.Measure (CoeffSpace d)} [MeasureTheory.IsFiniteMeasure P] {N : ℝ} (hN : 1 ≤ N) {F G : CoeffSpace d → BlockMat d} (hF : SchattenMemLp P N F) (hG : SchattenMemLp P N G) (hFpos : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (ofFullBlockMat 0) (F a)) (hFG : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (F a) (G a)) :
have MF := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (F a) α β ∂P); have MG := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (G a) α β ∂P); 0 ≤ blockTrace (blockSub MG MF) ∧ absSchattenNorm N (blockSub MG MF) ≤ blockTrace (blockSub MG MF) ∧ blockTrace (blockSub MG MF) ≤ 2 * ↑d * blockOpNorm MG ∧ blockTrace (blockSub MG MF) ≤ (2 * ↑d) ^ (1 - N⁻¹) * blockOpNorm MG ^ (1 - N⁻¹) * blockTrace (blockSub MG MF) ^ N⁻¹

Deterministic mean-gap bounds used in the centering step.

theorem HCPolySupport.HighContrast.Analysis.positiveGap_nonlinear {d : ℕ} {P : MeasureTheory.Measure (CoeffSpace d)} [MeasureTheory.IsProbabilityMeasure P] {N : ℝ} (hN : 1 < N) {F G : CoeffSpace d → BlockMat d} (hF : SchattenMemLp P N F) (hG : SchattenMemLp P N G) (hFpos : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (ofFullBlockMat 0) (F a)) (hFG : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (F a) (G a)) :
have MF := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (F a) α β ∂P); have MG := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (G a) α β ∂P); (lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (G a) (F a)) ≤ blockOpNorm MG ^ (1 - N⁻¹) * blockTrace (blockSub MG MF) ^ N⁻¹ + (2 * ↑d) ^ ((N - 1) / N ^ 2) * (lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (G a) MG) ^ (1 - N⁻¹) * (lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (G a) (F a)) ^ N⁻¹

l.fixed.geometry.positive.gap: the nonlinear inequality, with genuine scalar moments and finite real norm readbacks. The Hölder inputs are the centered majorant and gap Schatten functions, with conjugate exponents N/(N-1) and N.

theorem HCPolySupport.HighContrast.Analysis.positiveGap_young_absorption {d : ℕ} {P : MeasureTheory.Measure (CoeffSpace d)} [MeasureTheory.IsProbabilityMeasure P] {N : ℝ} (hN : 1 < N) {F G : CoeffSpace d → BlockMat d} (hF : SchattenMemLp P N F) (hG : SchattenMemLp P N G) (hFpos : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (ofFullBlockMat 0) (F a)) (hFG : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (F a) (G a)) :
have MF := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (F a) α β ∂P); have MG := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (G a) α β ∂P); (lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (G a) (F a)) ≤ ((2 * ↑d) ^ N⁻¹ * lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (G a) MG) + N / (N - 1) * blockOpNorm MG ^ (1 - N⁻¹) * blockTrace (blockSub MG MF) ^ N⁻¹

l.fixed.geometry.positive.gap: Young absorption of the proved nonlinear estimate, with the printed dimension and rational coefficients.

Pure block algebra identity behind the centering step.

theorem HCPolySupport.HighContrast.Analysis.positiveGap_centered_bound {d : ℕ} {P : MeasureTheory.Measure (CoeffSpace d)} [MeasureTheory.IsProbabilityMeasure P] {N : ℝ} (hN : 1 < N) {F G : CoeffSpace d → BlockMat d} (hF : SchattenMemLp P N F) (hG : SchattenMemLp P N G) (hFpos : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (ofFullBlockMat 0) (F a)) (hFG : ∀ᵐ (a : CoeffSpace d) ∂P, BlockMatLoewnerLE (F a) (G a)) :
have MF := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (F a) α β ∂P); have MG := ofFullBlockMat (Matrix.of fun (α β : BlockCoord d) => ∫ (a : CoeffSpace d), blockMatEntry (G a) α β ∂P); (lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (F a) MF) ≤ ((1 + (2 * ↑d) ^ N⁻¹) * lqSchattenNorm P N fun (a : CoeffSpace d) => blockSub (G a) MG) + (N / (N - 1) + (2 * ↑d) ^ (1 - N⁻¹)) * blockOpNorm MG ^ (1 - N⁻¹) * blockTrace (blockSub MG MF) ^ N⁻¹

l.fixed.geometry.positive.gap: center using two sharp subtraction inequalities, then bound the deterministic mean gap and insert Young absorption.