Tuned high-side scalar data: reserved loss, conductor cap and the record
DefinitionTransport_PhysicalHigh_TunedDataThroughout, , 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 .
This file fixes the scalar data of the high-side estimate at the tuned boundary.
-
Reserved loss. .
-
Conductor cap. , the largest conductor exponent that the tuned detector is allowed to reach.
-
Tuned data. For a natural number and a real allowance , a record consists of
- a positive even number of slots ;
- slot lengths , , pairwise distinct, with ;
- the minimal slot length and the minimal conductor ratio , with
- a system of detector scales for the parameters , that is, positive reals bounded by the allowance and satisfying the budget inequalities of the definition
Transport_PhysicalHigh_Parameters.
The record collects every scalar choice of the high-side argument in the order in which the source makes them: the loss and the mesh are fixed first, then and the slots, and the detector scales last, so that the allowance may depend on the final contradiction gap without affecting the slots.
Formalization Note Transcribed verbatim from the source. reservedLoss is an abbrev; the structure field P has type DetectorScales reservedLoss K rmin ellMin allowance. The source's derived inequalities about these objects (for example conductorCap_pos, TunedData.high_budget) are theorems and are not included here.
import Mathlib import Definitions.Def_Transport_PhysicalHigh_Parameters import Definitions.Def_Transport_PhysicalHigh_EnergyMesh set_option autoImplicit false /-! # Complete tuned high-side scalar data The reserved high loss and capacity are fixed first. The energy mesh is therefore fixed before K. The slot construction allows any prior lower bound on K. The positive allowance may depend on the later contradiction gap; it only affects post-slot errors, not the selected slots. -/ namespace RH.Transport.PhysicalHigh noncomputable section open scoped BigOperators abbrev reservedLoss : ℝ := 1 / 1000000000000000 noncomputable def conductorCap : ℝ := TunedCandidate.h + TunedCandidate.zeta + 2 * reservedLoss structure TunedData (Kmin : ℕ) (allowance : ℝ) where K : ℕ K_large : Kmin ≤ K K_pos : 0 < K K_even : Even K ell : Fin K → ℝ ellMin : ℝ rmin : ℝ ellMin_eq : ellMin = TunedCandidate.ell / (2 * K) rmin_eq : rmin = ellMin / conductorCap ellMin_pos : 0 < ellMin rmin_pos : 0 < rmin slots_injective : Function.Injective ell slots_sum : (∑ j, ell j) = TunedCandidate.ell slots_pos : ∀ j, 0 < ell j slots_lower : ∀ j, ellMin ≤ ell j slots_upper : ∀ j, ell j ≤ (1 / 200 : ℝ) * reservedLoss slots_fine : ∀ j, ell j ≤ plainMesh reservedLoss / 200 P : DetectorScales reservedLoss K rmin ellMin allowance end end RH.Transport.PhysicalHigh