Documentation

LeanPool.HighContrastHomogenization.Support.Probability.IndependentSums.GammaSigmaConcentration.Preliminaries

Coarse-graining support: Support.Probability.IndependentSums.GammaSigmaConcentration.Preliminaries #

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

Explicit upper bound for the Chapter 4 tail integral ∫_1^∞ t e^{-t^σ} dt. We use the full (0, ∞) Gamma-integral value because it is simpler to package and still sufficient for the generic heavy-tail backend.

Equations
Instances For

    Explicit logarithmic-control constant for the Γ_σ specialization.

    Equations
    Instances For

      Auxiliary union-term constant for the corrected small-σ Chapter 4 concentration split.

      Equations
      Instances For

        Explicit one-sided sqrt(card) scale for the heavy-tail Γ_σ concentration theorem on the range σ ∈ (0, 1).

        Equations
        Instances For

          Symmetric sqrt(card) scale for the heavy-tail Γ_σ concentration theorem on the range σ ∈ (0, 1). The factor 2^(1/σ) absorbs the two-sided tail union.

          Equations
          Instances For

            The concrete logarithmic control from the Chapter 4 notes, with an explicit choice of constant that is convenient in Lean.

            The Γ_σ tail integral on (1, ∞) is controlled by the explicit Gamma constant gammaSigmaTailIntegralConst σ.

            theorem HCPolySupport.IndependentSums.gammaSigma_log_constraint {σ l L t : ℝ} (hσ₀ : 0 < σ) (hσ₁ : σ < 1) (hlL : l ≤ 1 / 2 * L ^ (σ - 1)) (ht : t ∈ Set.Icc 1 L) :

            If λ ≤ (1/2) L^(σ - 1) and 1 ≤ t ≤ L, then the deterministic part of the Γ_σ logarithmic kernel constraint holds.

            theorem HCPolySupport.IndependentSums.smallRegime_sq_le_gammaSigmaTailPower {σ R t : ℝ} (hσ₀ : 0 < σ) (hσ₁ : σ < 1) (hR_pos : 0 < R) (ht_pos : 0 < t) (hsmall : t ≤ R ^ (σ / (2 * (2 - σ)))) :
            t ^ 2 ≤ (√R / t) ^ (σ / (1 - σ))
            theorem HCPolySupport.IndependentSums.smallRegime_cardPow_le_gammaSigmaTailPower {σ R t : ℝ} (hσ₀ : 0 < σ) (hσ₁ : σ < 1) (hR_pos : 0 < R) (ht_pos : 0 < t) (hsmall : t ≤ R ^ (σ / (2 * (2 - σ)))) :
            R ^ (σ / (2 - σ)) ≤ (√R / t) ^ (σ / (1 - σ))
            theorem HCPolySupport.IndependentSums.largeRegime_cardPow_le_gammaSigmaScale {σ R t : ℝ} (hσ₀ : 0 < σ) (hσ₁ : σ < 1) (hR_pos : 0 < R) (ht_pos : 0 < t) (hlarge : R ^ (σ / (2 * (2 - σ))) ≤ t) :
            R ^ (σ / (2 - σ)) ≤ R ^ (σ / 2) * t ^ σ
            theorem HCPolySupport.IndependentSums.largeRegime_correction_le_gammaSigmaScale {σ R t : ℝ} (hσ₀ : 0 < σ) (hσ₁ : σ < 1) (hR_pos : 0 < R) (ht_pos : 0 < t) (hlarge : R ^ (σ / (2 * (2 - σ))) ≤ t) :
            R ^ σ * t ^ (2 * σ - 2) ≤ R ^ (σ / 2) * t ^ σ