Theorem I.16: Schumacher code rate≤H(averageState)+Kd alpha/sqrt(n); Fe≥1−4N e^(−2mu²alpha²).
ProvedQAlgorithms.Papers.Independent.Renew46.P43.schumacherSchemeLet a finite pure-state iid source have average on a nonzero finite system of Hilbert dimension . Compression spaces and block lengths are positive, decoders are CPTP and all entropies/rates are in bits. The fixed-error rates are limsups of the actual block minimum of . Quantum encoding requires a CPTP extension; arbitrary encoding is state valued. In finite-block bounds , , and are the support rank and least positive eigenvalue of . The Schumacher code compresses onto the nonzero variance-typical range and decodes by inclusion; the strong converse permits any arbitrary-encoding code of fidelity error at most . For the intrinsic version the source letters and decoder lie in the specified classical, hybrid or full matrix algebra, with its intrinsic trace.
This is the source result with its stated operational or information-theoretic domain.
import Definitions.Def_QAlgorithms_Papers_Independent_Renew46_P43
noncomputable section
namespace QAlgorithms.Papers.Independent.Renew46.P43
open scoped BigOperators Matrix ComplexConjugate ComplexOrder ENNReal Matrix.Norms.L2Operator
open Filter MeasureTheory
open QAlgorithms.Papers.Independent.Information100.P08 (IsPure)
open QAlgorithms.Papers.Independent.Information100.P06 (Below)
open QAlgorithms.Papers.Independent.Information100.P60 (copies relativeBits)
open QAlgorithms.Papers.Independent.Dependency50.P19 (count both)
open QAlgorithms.Papers.Independent.Information100.P22 (coherentInformation)
attribute [local instance] Classical.propDecidable Classical.decEq
/-- Theorem I.16, PDF p.16: the actual range compression followed by inclusion.
The code's range is required nonzero: the printed encoder divides by dim K. -/
theorem schumacherScheme {a d : ℕ} (S : PureSource a d) (spectral : FineResolution d)
(haverage : spectral.state = S.average) (n : ℕ) (hn : 0 < n)
(alpha : ℝ) (halpha : 0 < alpha) (k : ℕ) (hk : 0 < k)
(V : Matrix (Fin n → Fin d) (Fin k) ℂ)
(hisometry : Vᴴ * V = 1) (hrange : V * Vᴴ = varianceProjector spectral n alpha) :
Real.logb 2 k / n ≤ vnEntropy S.average + varianceConstant * d * alpha / Real.sqrt n ∧
sourceFe (copies n S.average) (conjSuperop V ∘ projectionEncoder V) ≥
1 - 2 * varianceTail spectral alpha := by sorry
end QAlgorithms.Papers.Independent.Renew46.P43
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be a source of distinct pure density matrices on , with , , a probability mass function allowing zero probabilities, and average . Let a spectral resolution consist of nonnegative weights summing to one and a unitary matrix such that ; zero weights and repeated weights are allowed. For natural numbers with and , a real number , and a complex matrix , suppose and , where, writing for column of ,
and counts occurrences of in . Thus a zero spectral weight requires its occurrence count to be zero. Define to be the number of indices with , and let be the minimum of these positive numbers, or zero if there are none. Set , which is zero when , and . Then
where and . Here for , defined by continuous functional calculus, and , where , , and the amplification applies to each block on the second factor.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.