Sandwiched channel minimax with attained extrema
OpenQAlgorithms.Papers.Independent.Intakenext100.P15.sandwiched_channel_minimaxAll Hilbert spaces are finite-dimensional and nonzero. A density operator is positive semidefinite with trace one; singular states are included. All logarithms are base two. For real , denotes the sandwiched Rényi divergence and the traditional Petz divergence, with value on support failure and negative spectral powers taken on the support.
Write , where is unnormalized, and . Its reference marginal is and its channel-input marginal is . Define over all normalized densities. The finite attained extrema are retained as conclusions in the minimax statement.
For every CPTP and real ,
Both optimizations range over all density operators. There are a real and densities with and
for every density . Every fixed-input inner infimum and every fixed-comparison inner supremum also has an attaining witness, allowing infinite fixed-comparison suprema. Thus the displayed outer extrema and the original channel maximum are attained. This is the exchange of minimum and maximum used in channel additivity.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P15
noncomputable section
open scoped Matrix ComplexOrder ComplexConjugate Topology
open Filter
namespace QAlgorithms.Papers.Independent.Intakenext100.P15
attribute [local instance] Classical.propDecidable
attribute [local instance] Classical.decEq
-- These are pinned accepted definitions, not other chunks from the current run.
open QAlgorithms.Papers.Independent.Bridge50.P43 (NormalizedDensity payoff)
open QAlgorithms.Papers.Independent.Dependency50.P09 (Divergence logExt sandwichedD petzD)
open QAlgorithms.Papers.Independent.Information100.P34 (choi)
open QAlgorithms.Papers.Independent.Information100.P63 (schatten)
open QAlgorithms.Papers.Independent.Information100.P71
(supportPower SupportIncluded stateRelative channelPower diamond)
/-- PDF pp.11--12, Lemma 5, (5.16). Equality, finite value, saddle witnesses,
and attainment of every inner minimum/maximum elaborate the printed min/max. -/
theorem sandwiched_channel_minimax
{A B : Type} [Fintype A] [Fintype B] [DecidableEq A] [DecidableEq B]
[Nonempty A] [Nonempty B] (N : Superop A B) (hN : IsChannel N)
(alpha : ℝ) (ha : 1 < alpha) :
channelInfo (sandwichedD alpha) N =
(⨅ sigma : NormalizedDensity B, ⨆ rho : NormalizedDensity A,
sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma.1)) ∧
(∃ (v : ℝ) (rho0 : NormalizedDensity A) (sigma0 : NormalizedDensity B),
channelInfo (sandwichedD alpha) N = (v : EReal) ∧
sandwichedD alpha (eaChoiOutput N rho0.1) (Matrix.kronecker rho0.1 sigma0.1) = (v : EReal) ∧
(∀ rho : NormalizedDensity A,
sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma0.1) ≤ (v : EReal)) ∧
(∀ sigma : NormalizedDensity B,
(v : EReal) ≤ sandwichedD alpha (eaChoiOutput N rho0.1) (Matrix.kronecker rho0.1 sigma.1))) ∧
(∀ rho : NormalizedDensity A, ∃ sigma : NormalizedDensity B,
sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma.1) =
inputInfo (sandwichedD alpha) N rho) ∧
(∀ sigma : NormalizedDensity B, ∃ rho0 : NormalizedDensity A,
sandwichedD alpha (eaChoiOutput N rho0.1) (Matrix.kronecker rho0.1 sigma.1) =
⨆ rho : NormalizedDensity A,
sandwichedD alpha (eaChoiOutput N rho.1) (Matrix.kronecker rho.1 sigma.1)) := by sorry
end QAlgorithms.Papers.Independent.Intakenext100.P15
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be nonempty finite types with decidable equality, let for a finite family satisfying , and fix a real . Let denote the positive semidefinite trace-one matrices on , define by continuous functional calculus from the function equal to zero at zero and elsewhere, set , and put . Define for , for , and otherwise, and set
All real arguments and factors here are embedded into the extended reals. Writing , the theorem asserts
and the existence of a real , , and such that , for every , and for every . It also asserts that for every fixed some attains , and for every fixed some attains . The saddle value is explicitly finite because is real; the other attained values may be extended real, and singular density matrices are allowed.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.