Documentation

LeanPool.HighContrastHomogenization.Support.Book.Ch04.Theorems.PartitionAverageMoments.Rosenthal

Coarse-graining support: Support.Book.Ch04.Theorems.PartitionAverageMoments.Rosenthal #

Imported from the Apache-2.0 CoarseGraining development at commit c7ddd76c08ade64fed1b8d2ca51be14dfee8deb4.

Restriction-local partition-average moment estimates: Rosenthal bounds #

theorem HCPolySupport.Book.Ch04.integral_abs_finiteSum_pow_rpow_inv_le_rosenthal_descendantScaleColorClass {d : ℕ} {Q : TriadicCube d} {k : ℤ} {c : ScaleColor d k} {P : RestrictionCoeffLaw d} [MeasureTheory.IsProbabilityMeasure P] {p : ℕ} {K : ℝ} (hP : RestrictionUnitRangeDependentLaw P) (hp : 2 ≤ p) (hK_nonneg : 0 ≤ K) (X : TriadicCube d → RegCoeffField d → ℝ) (hX_local : ∀ R ∈ descendantsAtScaleScaleColorClass Q k c, IsRestrictionLocalRandomVariable (cubeSet R) ⋯ (X R)) (hX_aemeas : ∀ R ∈ descendantsAtScaleScaleColorClass Q k c, AEMeasurable (X R) P) (hLp_int : ∀ R ∈ descendantsAtScaleScaleColorClass Q k c, MeasureTheory.Integrable (fun (a : RegCoeffField d) => |X R a| ^ p) P) (h_mean : ∀ R ∈ descendantsAtScaleScaleColorClass Q k c, ∫ (a : RegCoeffField d), X R a ∂P = 0) (hK : ∀ R ∈ descendantsAtScaleScaleColorClass Q k c, (∫ (a : RegCoeffField d), |X R a| ^ p ∂P) ^ (1 / ↑p) ≤ K) :
(∫ (a : RegCoeffField d), |∑ R ∈ descendantsAtScaleScaleColorClass Q k c, X R a| ^ p ∂P) ^ (1 / ↑p) ≤ 2 * ↑p * (↑(descendantsAtScaleScaleColorClass Q k c).card ^ (1 / ↑p) * K) + 4 * rosenthalBennettIntegralConst * (√↑p * (√↑(descendantsAtScaleScaleColorClass Q k c).card * K))

Rosenthal's L^p bound on a single scale-color class of descendants under the restriction-unit-range and restriction-local-random-variable assumptions.

theorem HCPolySupport.Book.Ch04.integral_abs_finiteSum_pow_rpow_inv_le_rosenthal_uniform_descendants {d : ℕ} {Q : TriadicCube d} {k : ℤ} {P : RestrictionCoeffLaw d} [MeasureTheory.IsProbabilityMeasure P] {p : ℕ} {K : ℝ} (hP : RestrictionUnitRangeDependentLaw P) (hp : 2 ≤ p) (hK_nonneg : 0 ≤ K) (X : TriadicCube d → RegCoeffField d → ℝ) (hX_local : ∀ R ∈ descendantsAtScale Q k, IsRestrictionLocalRandomVariable (cubeSet R) ⋯ (X R)) (hX_aemeas : ∀ R ∈ descendantsAtScale Q k, AEMeasurable (X R) P) (hLp_int : ∀ R ∈ descendantsAtScale Q k, MeasureTheory.Integrable (fun (a : RegCoeffField d) => |X R a| ^ p) P) (h_mean : ∀ R ∈ descendantsAtScale Q k, ∫ (a : RegCoeffField d), X R a ∂P = 0) (hK : ∀ R ∈ descendantsAtScale Q k, (∫ (a : RegCoeffField d), |X R a| ^ p ∂P) ^ (1 / ↑p) ≤ K) :

Rosenthal's L^p bound for sums over all descendants at a fixed scale, using the restriction-local and restriction-unit-range interfaces.