Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section4.Provider.BoundsEaL.PerCubeBracket

PerCubeBracket #

Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.

Step 5's (1 + ∇k · λ^{-1}) bracket, at every moment #

What this module does #

Two of the three summands of MajorantSlots.step3DisplayAt carry the bracket

( 1 + 3^{2j}‖∇(k_L − k_{j−2})‖_{W^{1,∞}(□_j)} · λ_{2γ,2}^{-1} )^{θ} ,

at the two printed exponents θ = 2s/(1−4γ) and θ = 4γ/(1−4γ) (the development's gapped-gauge reading of the printed 2s/(1−2γ), 2γ/(1−2γ); Step3HSlot's deviation, inherited). This module supplies its q-th moment, for every q ∈ [1,∞), in the development normal form ∫⁻ (·)^q ≤ (ofReal R)^q.

step3Bracket_majorant_le then evaluates the product majorant: the whole q-th-moment bound is

1 + C(2)√(2q) · fullGradConst M · (3^{γ j} σ_{j−1}^{-1})
      · ( λUp·C_cg + C(1/3)(2q)³ · λUp·λMax · cgTailScale ) ,

and the middle factor is O(√γ) by bullet (B3) together with the σ index reconciliation of SigmaBarLandmark, which is exactly the statement that the bracket is 1 + O(√q √γ) in the printed regime.

1. The abstract bracket moment #

theorem Algsuperdiff.Section4.Provider.BoundsEaL.lintegral_rpow_one_add_mul_le_of_moments {Omega : Type u_1} [MeasurableSpace Omega] {mu : MeasureTheory.Measure Omega} [MeasureTheory.IsProbabilityMeasure mu] {X Y : Omega → ℝ} {RX RY q theta : ℝ} (hq : 1 ≤ q) (hth1 : theta ≤ 1) (hX0 : ∀ (omega : Omega), 0 ≤ X omega) (hY0 : ∀ (omega : Omega), 0 ≤ Y omega) (hXm : AEMeasurable X mu) (hYm : AEMeasurable Y mu) (hRX : 0 ≤ RX) (hRY : 0 ≤ RY) (hX : ∫⁻ (omega : Omega), ENNReal.ofReal (X omega) ^ (2 * q) ∂mu ≤ ENNReal.ofReal RX ^ (2 * q)) (hY : ∫⁻ (omega : Omega), ENNReal.ofReal (Y omega) ^ (2 * q) ∂mu ≤ ENNReal.ofReal RY ^ (2 * q)) :
∫⁻ (omega : Omega), ENNReal.ofReal ((1 + X omega * Y omega).rpow theta) ^ q ∂mu ≤ ENNReal.ofReal (1 + RX * RY) ^ q

The bracket moment, over an abstract probability space.

If two nonnegative observables satisfy the development normal form at the doubled exponent 2q, then for every θ ≤ 1 the bracket (1 + X·Y)^θ satisfies it at q, with the majorant 1 + R_X R_Y.

The θ ≤ 1 step is x^θ ≤ x for x ≥ 1: no moment input is spent on the exponent. The product step is the two-factor Hölder engine and the additive 1 is ℝ≥0∞ Minkowski against a constant, which a probability measure integrates to 1.

2. Both printed exponents are at most one #

theorem Algsuperdiff.Section4.Provider.BoundsEaL.step3_gradient_exponent_le_one {gam s : ℝ} (hgam : gam ≤ 1 / 8) (hs : s ≤ 1 / 4) :
2 * s / (1 - 4 * gam) ≤ 1

The gradient bracket's exponent. 2s/(1−4γ) ≤ 1 in the anchor's own ranges s ≤ 1/4, γ ≤ 1/8. This is the exact endpoint: at s = 1/4 and γ = 0 the exponent IS 1.

theorem Algsuperdiff.Section4.Provider.BoundsEaL.step3_gamma_exponent_le_one {gam : ℝ} (hgam : gam ≤ 1 / 8) :
4 * gam / (1 - 4 * gam) ≤ 1

The γ bracket's exponent. 4γ/(1−4γ) ≤ 1 at γ ≤ 1/8.

3. The two slots are nonnegative #

The L-free gradient slot is nonnegative at the canonical tail gauge.

The λ-slot of Step 3's display is measurable: the gapped gauge 2γ is positive, which is all MajorantMeasurability's literal-measurability engine needs.

4. The bracket at the Step-3 slots #

Step 5's bracket, at every moment.

For every q ∈ [1,∞) and every exponent θ ≤ 1 — in particular the two printed exponents 2s/(1−4γ) and 4γ/(1−4γ), by step3_gradient_exponent_le_one and step3_gamma_exponent_le_one — the bracket of MajorantSlots.step3DisplayAt obeys the development normal form with the majorant 1 + R_{B6a}(2q) · R_{B5}(2q), the product of the two proved Step-4 majorants at the doubled exponent.

The binders are the printed regime (bullet (B5)'s own), the range R.scale ≤ m of the gradient bullet, and 1 ≤ q. No γ-power and no s-power is moved.

5. The bracket majorant is 1 + O(√q √γ) #