LocalizationOscillationGridMoment #
Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.
Provider: the grid fourth-moment functional of e.lower.bound.oscillations #
Source displays in ABK26:
e.lower.bound.oscillations(label; display) opens with the quantity(avsum_{z in 3^n Zd cap cu_K} E[ ‖grad w - (grad w)_{z+cu_n}‖^4_{L2bar(z+cu_n)} ])^{1/4};e.nablaw.oscillations(label; display) is the per-base-point estimate that gets inserted into it.
This module isolates the purely functional-analytic bookkeeping of that passage: the mixed "average over the grid, expectation over the sample" fourth-moment functional and its behaviour under a pointwise two-term majorant.
What is proved #
gridFourthMoment,gridFourthMomentRoot-- the functionalavsum_{R in I} integral^4of the display and its fourth root.cubeFamilyAverage_mono,cubeFamilyAverage_add,cubeFamilyAverage_const_mul-- the three elementary facts about the grid average ofLocalizationGridthat the assembly uses.gridFourthMoment_le_eight_mul_add-- ifF <= G + Hpointwise on the grid and on the sample space, thenM_4(F) <= 8 M_4(G) + 8 M_4(H). The elementary inequality behind it is(a+b)^4 <= 8(a^4+b^4), valid for all reals.gridFourthMomentRoot_le_add_of_le-- the shape the display consumes: fromM_4(G) <= WG^4andM_4(H) <= WH^4one getsM_4(F)^{1/4} <= 8^{1/4} (WG + WH).gridFourthMoment_le_of_forall_le-- a per-cube uniform bound transfers to the grid average.
Divergences from the printed statement #
- The triangle step costs the absolute factor
8^{1/4}, not1. The manuscript's implicit step is Minkowski's inequality inL^4(counting x P), whose constant is1. What is proved here is the elementary(a+b)^4 <= 8(a^4+b^4)route, whose constant is8^{1/4} < 1.682. The loss is an absolute number, it is written out in the statement, and it is absorbed by the manuscript's dimension-onlyC; no sharper claim is made and nothing is hidden inside a constant. - Nothing here relates the full grid to its interior part. The passage
from an interior mesh to the full one, and the boundary strip it leaves
behind, are
LocalizationFluctuationMeshTransfer.cubeFamilyAverage_le_add_sqrt_boundaryFractionand its consumers; this module works at a fixed familyIthroughout.
Two elementary real facts #
(a + b)^4 ≤ 8 (a^4 + b^4), valid for all reals. This is the elementary
substitute for Minkowski's inequality used in the grid assembly; the constant
8 is sharp for this form.
Three elementary facts about the grid average #
The grid fourth-moment functional #
The fourth-moment functional of e.lower.bound.oscillations. For a family
F of nonnegative random variables indexed by the grid I, this is avsum_{R in I} E[^4], the quantity under the outer fourth root of the display.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The outer fourth root of the display.
Equations
Instances For
Scalars come out of the functional at the fourth power.
A uniform per-cube bound transfers to the grid average.
The two-term majorant #
The grid fourth moment under a pointwise two-term majorant. This is the
step of e.lower.bound.oscillations that inserts e.nablaw.oscillations into the
grid average, with the elementary constant 8 in place of Minkowski's 1.
From a fourth-power bound on the functional to a bound on its fourth root.
The shape e.lower.bound.oscillations consumes. A pointwise two-term
majorant on the grid, together with fourth-moment bounds for the two majorants,
gives the fourth root of the display's left-hand side at the absolute cost
8^{1/4}.