Unnumbered conditional-entropy rederivation
OpenQAlgorithms.Papers.Independent.Support50.P14.gaussianConditionalEntropyLet and be nonempty finite collections of full oscillator modes, and let be a normalized density with finite complete moments. Let be a Gaussian density with the same symmetric covariance, allowing different first moments. All four entropies of the joint states and their marginals are finite. With measured in nats,
Thus Gaussian states maximize conditional entropy at fixed covariance. The sign is opposite to the functional used by Lemma 1.
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P14
noncomputable section
open Filter
open scoped BigOperators ENNReal Topology
attribute [local instance] Classical.propDecidable Classical.decEq
namespace QAlgorithms.Papers.Independent.Support50.P14
open QAlgorithms.RealRegisters (Hilbert)
open QAlgorithms.Papers.Independent.Dependency50.P33
(State IsDensity IsTraceClass traceSize)
open QAlgorithms.Papers.Independent.Dependency50.P39 (displacementMatrix)
open QAlgorithms.Papers.Independent.Frontier30.P14
(unitaryCoeff conjugateKernel boundedUnitaryConjugation)
open QAlgorithms.Papers.Independent.Frontier18.P07 (extendedEntropyNats)
open QAlgorithms.Papers.Independent.Claimed50.P47 (Occupation)
open QAlgorithms.Papers.Independent.Gap48.P20 (Copies)
/-- Conditional entropy rederived on PDF p.2: S(AB)-S(A) is MAXIMIZED by the
Gaussian counterpart. All four entropies are finite; toReal never hides infinity.
The input and comparator means may differ by local displacements. -/
theorem gaussianConditionalEntropy {A B : Type} [Fintype A] [Fintype B]
[Nonempty A] [Nonempty B] (rho rhoG : State (Occupation (A ⊕ B)))
(d dG : InterleavedQuadrature (A ⊕ B) → ℝ) (gamma : OscillatorCovariance (A ⊕ B))
(hmoments : CompleteMoments rho.coefficients d gamma)
(hgaussian : GaussianAt rhoG.coefficients dG gamma) :
extendedEntropyNats rho.coefficients < ⊤ ∧
extendedEntropyNats rhoG.coefficients < ⊤ ∧
extendedEntropyNats (marginalFirst rho.coefficients) < ⊤ ∧
extendedEntropyNats (marginalFirst rhoG.coefficients) < ⊤ ∧
(extendedEntropyNats rho.coefficients).toReal -
(extendedEntropyNats (marginalFirst rho.coefficients)).toReal ≤
(extendedEntropyNats rhoG.coefficients).toReal -
(extendedEntropyNats (marginalFirst rhoG.coefficients)).toReal := by sorry
end QAlgorithms.Papers.Independent.Support50.P14
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be finite, nonempty types, set , , and , and let be states with coefficient kernels on . A state kernel here admits an expansion with and , has positive semidefinite finite principal submatrices, and has summable real diagonal of sum . Let and let be a real matrix on . Define
with and the energy sum valued in . Assume , , and, for every , summability of each inner series in and of the outer series of inner sums; also assume that has derivative at , and its derivative function has derivative there. Assume these same conditions for with , together with
No equality of and is required. Define the first marginal of any kernel by
For a kernel on any label set, define
where is its finite principal compression, the logarithm is the real continuous functional calculus appearing in the code, scalar , and the empty compression contributes . The conclusion asserts finiteness of all four quantities , and the inequality
The subtraction is performed after converting these extended nonnegative entropies to real numbers; their asserted finiteness makes those conversions finite.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.