Average ensemble error and existence of a good frozen vector
OpenQAlgorithms.Papers.Independent.Intakethird100.P14.ensemble_error_boundLet be any binary-input classical–quantum channel with positive trace-one density outputs, let , and let be an arbitrary information set, with . Encode by and use the prescribed collective successive cancellation decoder: square-root-density spectral tests at information positions, identity at frozen positions, and uniform unknown future bits. Write for block error averaged uniformly over information messages with frozen vector , and for its further uniform average over . Then
The fidelity is squared fidelity. The result requires no polar selection rule and includes , when every bit is frozen and block error is zero. It transfers split-channel reliability to an actual sequential decoder.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P14
noncomputable section
namespace QAlgorithms.Papers.Independent.Intakethird100.P14
open Filter MeasureTheory
open scoped ENNReal NNReal ComplexOrder BigOperators
open QAlgorithms.RealRegisters (Hilbert)
open QAlgorithms.Papers.Independent.Dependency50.P33 (IsDensity)
open QAlgorithms.Papers.Independent.Dependency50.P13 (compression)
open QAlgorithms.Papers.Independent.Renew46.P44 (wordOutput)
attribute [local instance] Classical.propDecidable Classical.decEq
/-- Proposition 4, PDF p.6; proof p.9. Both the frozen-vector average and
existence of a frozen vector are conclusions, for every information set.
The factor 2 is OUTSIDE sqrt and the factor 1/2 is INSIDE the sum. -/
theorem ensemble_error_bound {I : Type} (W : Bit → OutputCoefficients I) (hW : IsBinaryCqChannel W)
(n : ℕ) (A : Finset (PolarWire n)) :
ensembleError W n A ≤
2 * Real.sqrt (∑ i ∈ A, (1 / 2 : ℝ) * Real.sqrt (squaredCqFidelity (splitChannel W n i))) ∧
∃ f : Frozen n A, frozenError W n A f ≤
2 * Real.sqrt (∑ i ∈ A, (1 / 2 : ℝ) * Real.sqrt (squaredCqFidelity (splitChannel W n i))) := by sorry
end QAlgorithms.Papers.Independent.Intakethird100.P14
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be any type, let , and let assign a complex coefficient array on to each bit . Assume each is a density array in the following precise sense: it has an expansion with and ; every quadratic form restricted to a finite subset of is real and nonnegative; and its real diagonal is summable with sum . Fix any , put , and let . Define over , where , , and reverses the binary digits of an index. For , set and , for . For an index , prefix , and bit , let , where the filled word equals for , equals at , and equals for . The split array , indexed by , is . For any coefficient array , let be a chosen bounded operator having coefficients in the coordinate basis if one exists, and otherwise; let be a chosen positive bounded operator satisfying if one exists, and otherwise; and put . For a binary array , define . For decoding, let be the orthogonal projection onto the closed span of all eigenvectors with negative real eigenvalue of , and set , . Given , form by processing , left multiplying by when and by the identity otherwise, and set . For frozen bits , define , and define . Then
This includes , when , and or ; an assignment on an empty set has one possible value. The basis traces use the total infinite-sum operation, which is for a nonsummable family.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.