Documentation

LeanPool.Superdiffusion.Support.Probability.EfronStein.Transfer

Transfer #

Ported from scottnarmstrong/CoarseGraining at c7ddd76c08ade64fed1b8d2ca51be14dfee8deb4. Module imports, support namespace and local option directives adapted for Lean Pool.

Efron–Stein transfer to coefficient-field laws #

This file transfers the abstract product-space Efron–Stein inequality (Homogenization.efronStein_pi) to a single restriction-unit-range-dependent probability measure P on the honest-fields carrier RegCoeffField d, resampled through a family of restriction endomorphisms.

Fix a Fintype ι and a family of pairwise AreUnitSeparated measurable regions C : ι → Set (Vec d) (the MeasurableSet side-conditions make the carrier restriction σ-algebra RestrictionSigmaR well defined). Writing R a i := restrictReg (C i) (hC i) a for the joint restriction map, each coordinate is a restriction-local random variable for RestrictionSigmaR (C i) (hC i), so the family is independent under a restriction-unit-range-dependent P and the pushforward P.map R factors as the product measure Measure.pi (fun i => P.map (restrictReg (C i) (hC i))). Efron–Stein on that product, transported back through the map identity, yields the variance bound for a bounded measurable observable G of the restricted fields.

The single resampling coordinate is restrictReg (C i) (hC i) a': an independent copy of P re-drawn only on C i.

Reference: the paper (Armstrong–Kuusi–Loher, to appear).

noncomputable def HCPolySupport.restrictObservable {d : ℕ} (U : Set (Vec d)) (hU : MeasurableSet U) :

The restriction endomorphism restrictReg U hU, bundled as the identity observable on the U-restricted field. This is the family fed to the carrier independence bridge in the Efron–Stein transfer.

Equations
Instances For

    The restriction observable is (globally) measurable on the carrier.

    The restriction observable is a restriction-local random variable on its observation set: it is measurable for the carrier restriction σ-algebra RestrictionSigmaR U hU.

    theorem HCPolySupport.efronStein_transfer_restriction {d : ℕ} {ι : Type u_1} [Fintype ι] [DecidableEq ι] {C : ι → Set (Vec d)} (hC : ∀ (i : ι), MeasurableSet (C i)) (hsep : Pairwise fun (i j : ι) => AreUnitSeparated (C i) (C j)) {P : MeasureTheory.Measure (RegCoeffField d)} [MeasureTheory.IsProbabilityMeasure P] (hP : IsRestrictionUnitRangeDependentR P) {G : (ι → RegCoeffField d) → ℝ} (hG : Measurable G) {M : ℝ} (hMG : ∀ (x : ι → RegCoeffField d), |G x| ≤ M) (R : RegCoeffField d → ι → RegCoeffField d) (hRdef : R = fun (a : RegCoeffField d) (i : ι) => restrictReg (C i) ⋯ a) :
    ProbabilityTheory.variance (G ∘ R) P ≤ 1 / 2 * ∑ i : ι, ∫ (a : RegCoeffField d), ∫ (a' : RegCoeffField d), (G (Function.update (R a) i (restrictReg (C i) ⋯ a')) - G (R a)) ^ 2 ∂P ∂P

    Restriction Efron–Stein transfer. For a finite family of pairwise AreUnitSeparated measurable regions C i, a restriction-unit-range-dependent probability measure P on the carrier RegCoeffField d, and a bounded measurable observable G of the jointly restricted fields R a = fun i => restrictReg (C i) (hC i) a, the variance of G ∘ R is controlled by the sum of single-region resampling energies, each an independent copy of P re-drawn only on C i.