Documentation

LeanPool.HighContrastHomogenization.Support.Book.Ch05.Theorems.Section57.BadScaleComponentRows

Coarse-graining support: Support.Book.Ch05.Theorems.Section57.BadScaleComponentRows #

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

Row estimates for split bad-scale components #

The lemmas in this file are deliberately small. They convert a fixed-pair tail bound, a deterministic lower bound on the tail parameter, and a descendant-cardinality estimate into the weighted row estimates needed by the component summation lemmas.

theorem HCPolySupport.Book.Ch05.Section57.measureReal_highBottomPairEvent_le_weighted_row_of_badPair_bound {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} [MeasureTheory.IsFiniteMeasure μ] {H : ℕ → ℕ → Ω → ℝ} {K a t α : ℝ} {q m n r : ℕ} {S D A ρ τ lam w : ℝ} (hS : 0 ≤ S) (hw : 0 ≤ w) (hD : D ≤ w ^ q * w ^ r) (hAρ_nonneg : 0 ≤ A * ρ ^ r) (hlam : A * ρ ^ r ≤ lam) (hτ : 0 < τ) (hbad : μ.real (badPairEvent H t α q m n) ≤ S * (D * Real.exp (-lam ^ τ))) :
μ.real (highBottomPairEvent H K a t α q m n) ≤ S * w ^ q * (w ^ r * Real.exp (-(A * ρ ^ r) ^ τ))

High-bottom row estimate from a fixed-pair no-log bound and deterministic weight estimates.

theorem HCPolySupport.Book.Ch05.Section57.measureReal_crudeBottomPairEvent_le_weighted_row_of_badPair_bound {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} [MeasureTheory.IsFiniteMeasure μ] {H : ℕ → ℕ → Ω → ℝ} {K a t α : ℝ} {q m n r : ℕ} {S D A ρ σ lam w : ℝ} (hS : 0 ≤ S) (hw : 0 ≤ w) (hD : D ≤ w ^ q * w ^ r) (hAρ_nonneg : 0 ≤ A * ρ ^ r) (hlam : A * ρ ^ r ≤ lam) (hσ : 0 < σ) (hbad : μ.real (badPairEvent H t α q m n) ≤ S * (D * Real.exp (-lam ^ σ))) :
μ.real (crudeBottomPairEvent H K a t α q m n) ≤ S * w ^ q * (w ^ r * Real.exp (-(A * ρ ^ r) ^ σ))

Crude-bottom row estimate from a fixed-pair crude bound and deterministic weight estimates.

theorem HCPolySupport.Book.Ch05.Section57.descendantsAtScale_bottom_row_card_le_weight {d N q r : ℕ} {j : Fin (q + 1)} (hnm : q - ↑j ≤ q + r) :
have D := descendantsAtScale (originCube d ↑(N + (q + r))) ↑(N + (q - ↑j)); have w := ↑(3 ^ d); ↑D.card ≤ w ^ q * w ^ r

Descendant-cardinality weight for bottom rows. In the range n = q - j, m = q + r, the number of descendants is bounded by (3^d)^q (3^d)^r.