Chosen moments: a height exponent for which the tuned source moment bound holds eventually
ProvedRH.Transport.PhysicalHigh.data_chosen_momentsThroughout, , a Hecke character is a finite-order Hecke character of in OpenAI's family-003 development, and is the supremum of the real parts of the zeros of the associated Hecke -functions (pole exception excluded, sentinel at ). The tuned constants of LW56e2's refinement are , the Gram exponent , the total slot length , the row and column exponents and , the conductor exponent , the cushion , the boundary and the signal exponent .
Let be source data, tuned data with detector scales , and counts count parameters for the modulus of at the reserved loss . Assume
Then there is a height exponent with
such that for every Hecke character there is a constant for which, for all sufficiently large ,
holds, where is the tuned source moment bound.
This is the moment input of the high-side estimate: the actual moments of the detector batches on all row subsets of the central partition are bounded with constant up to height . The ceiling is OpenAI's theorem and is a hypothesis here.
Formalization Note is TunedSourceMomentBound from the definition Transport_PhysicalHigh_MomentBound; the constant depends on , while does not.
import Mathlib import Definitions.Def_HeckeSevenEighths import Definitions.Def_OAIHecke78B027 import Definitions.Def_Transport_PhysicalHigh_MomentBound set_option autoImplicit false open RH RH.Transport RH.Transport.PhysicalHigh open scoped Classical open Filter open OAI OAI.SevenEighths open HeckeFamily ProbeFinalAssembly ProbeHighRowFamily HeckeInverseAmplification noncomputable section
theorem RH.Transport.PhysicalHigh.data_chosen_moments {Δ : ℝ} {seed : Parameters.HighData Δ} (F : SourceData seed)
{Kmin : ℕ} {allowance : ℝ} (D : TunedData Kmin allowance)
(counts : CountParameters F.modulus ⊤ reservedLoss)
(hbeta : (51/100:ℝ)≤HeckeZeroSupremum.beta)
(hceiling : HeckeZeroSupremum.beta≤7/8) :
∃τ : ℝ,0<τ ∧ τ<(1/200:ℝ)/2 ∧ 4*τ<(1/200:ℝ)*D.P.cost ∧
τ<reservedLoss ∧ 2*τ≤reservedLoss ∧ τ*(2+4*D.P.eps)<reservedLoss ∧
∀η : Character,∃C : ℝ,0<C ∧ ∀ᶠZ : ℝ in atTop,
1<Z ∧ TunedSourceMomentBound F D counts η Z τ (C*Z^reservedLoss) (Z^(2*τ)) := by sorry