Documentation

LeanPool.HighContrastHomogenization.Provider.Quenched.ScaledMaximumStrictSource

High-contrast homogenization: Provider.Quenched.ScaledMaximumStrictSource #

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

A scaled-maximum tail with a strict source event #

The source-tail assumption in CoarseEllipticityDagger is strict. This variant uses the strict slack cSrc < C from the half-threshold argument instead of strengthening that input to a closed tail event.

theorem HCPolySupport.HighContrast.Quenched.measureReal_scaled_max_one_tail_le_of_source_lt {Ω : Type u_1} [MeasurableSpace Ω] {P : MeasureTheory.Measure Ω} [MeasureTheory.IsFiniteMeasure P] {Rmix Ssrc : Ω → ℝ} {L C cMix cSrc cd η t : ℝ} {Ψ : ℝ → ℝ} (hL : 0 < L) (ht : 1 ≤ t) (hC : 1 < C) (hcMix : cMix ≤ C) (hcSrc : cSrc < C) (hmix : P.real {ω : Ω | cMix * t ≤ Rmix ω} ≤ Real.exp (-cd * t ^ η)) (hsrc : P.real (IndependentSums.upperTailEvent Ssrc (cSrc * t)) ≤ (Ψ (cSrc * t))⁻¹) :
P.real {ω : Ω | C * L * t ≤ L * max 1 (max (Rmix ω) (Ssrc ω))} ≤ Real.exp (-cd * t ^ η) + (Ψ (cSrc * t))⁻¹

Union bound for a scaled normalized maximum, with a closed mixing tail and a strict source tail. No independence between the random scales is needed.