Theorem II.3: total correlation equals quantum mutual information (original seed)
OpenQAlgorithms.Papers.Independent.Intakenext100.P47.totalErasure_eq_mutualInformationLet be any density matrix on finite nonzero complex spaces . For and , minimize each noise cost over ensembles of product unitaries whose output on is within full trace norm of a normalized product state. The allowed ensembles may be coordinated, independent across the parties, restricted to Alice, or restricted to Bob. The three costs are , the Shannon entropy , and entropy exchange . All finite minima exist and are bounded below; their normalized sequences are bounded above and below. Define by applying or to the normalized minima and then taking the supremum over . The outer sets are bounded above. For every choice of noise class , cost , and block limit ,
All entropies are in bits. This gives the operational meaning of total correlation and retains both displayed extreme equalities and the following robustness assertion.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P47
noncomputable section
namespace QAlgorithms.Papers.Independent.Intakenext100.P47
open QAlgorithms Filter
open QAlgorithms.Papers.Independent.Gap48.P20 (Copies localProduct)
open QAlgorithms.Papers.Independent.Dependency50.P19 (mutualInformation first)
open QAlgorithms.Papers.Independent.Dependency50.P45 (entropyExchange)
open QAlgorithms.Papers.Independent.Information100.P27 (shannon Config)
open QAlgorithms.Papers.Independent.Information100.P71 (channelPower)
open QAlgorithms.Papers.Independent.Advance50.P05 (RectangularInstrument)
open scoped Matrix ComplexConjugate
attribute [local instance] Classical.propDecidable Classical.decEq
/-- PDF p.4, Theorem II.3, displayed extremes and robustness paragraph.
The minima and bounds exclude totalized-real infimum/limit artifacts. -/
theorem totalErasure_eq_mutualInformation {A B : Type} [Fintype A] [Fintype B]
[Nonempty A] [Nonempty B] (rho : Matrix (A × B) (A × B) ℂ)
(hrho : IsDensityMatrix rho) :
(∀ eps > (0 : ℝ), ∀ mode cost,
(∀ n, 0 < n → (totalCosts rho n eps mode cost).Nonempty ∧
BddBelow (totalCosts rho n eps mode cost) ∧
sInf (totalCosts rho n eps mode cost) ∈ totalCosts rho n eps mode cost) ∧
BddBelow (Set.range (totalBlockRate rho eps mode cost)) ∧
BddAbove (Set.range (totalBlockRate rho eps mode cost))) ∧
(∀ mode cost pessimistic,
BddAbove {v | ∃ eps > (0 : ℝ),
v = rateLimit pessimistic (totalBlockRate rho eps mode cost)} ∧
totalErasure rho mode cost pessimistic = mutualInformation rho) := by sorry
end QAlgorithms.Papers.Independent.Intakenext100.P47
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For any finite nonempty types and any positive semidefinite matrix on with trace , define on by . A permitted random operation has a positive finite number of terms, nonnegative weights summing to , local unitaries , and action . There are four choices of restriction: no further restriction; for every ; for every ; or an indexing of the terms by pairs of local choices such that both the weights and the unitaries factor into independent local choices. There are three choices of cost: ; ; or the entropy exchange , where the canonical purification has coefficients and . Let be the set of these costs for operations satisfying for some density matrices , where . Put . The declaration asserts that, for every , every restriction and every cost, each set with is nonempty, bounded below, and contains its infimum, and the entire sequence , , is bounded both below and above. Furthermore, for every restriction, cost, and choice or , the set is bounded above and
Here the partial traces sum over the discarded index. The supremum ranges over all positive tolerances; the finite-block assertions exclude .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.