Capacity of an arbitrary channel sequence
OpenQAlgorithms.Papers.Independent.Support50.P20.arbitraryChannel_capacityFix a complex output Hilbert space . For each positive integer , let be an arbitrary channel from a nonempty set to normalized trace-class states on the completed tensor product . No relation between different blocklengths is assumed, and may include entangled block preparations. Define independently as the extended supremum of over sequences of actual -message codes whose average errors tend to zero. For each discrete law on , form and its marginals. Then the following outer limit exists, possibly with value , and
This operational formula covers channel sequences without causality, memorylessness, or stationarity. The small-error limit is taken after the lower limit in blocklength.
Formalization Note Completed tensor powers use square-summable coordinates indexed by tensor-product basis tuples. Sequence index zero denotes the one-use channel, so every blocklength denominator is positive. The equality with the outer limit is expressed as convergence to the independently defined operational capacity.
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P20
noncomputable section
open Filter
open scoped ENNReal
namespace QAlgorithms.Papers.Independent.Support50.P20
attribute [local instance] Classical.propDecidable Classical.decEq
open QAlgorithms.Papers.Independent.Dependency50.P13
(IsTraceClass IsPositive operatorTrace State)
open QAlgorithms.Papers.Independent.Information100.P63 (logTwo)
variable {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H]
/-- PDF pp.4–5, (10) and footnote [23]: arbitrary-channel sequence capacity.
The inner limit is liminf; epsilon decreases to zero AFTER it. Output coordinates
are those of B tensor (n+1), and inputs X_n are unrestricted block preparations.
The Tendsto formulation retains the outer limit and its existence, even at infinity. -/
theorem arbitraryChannel_capacity {I : Type*} [Nonempty I] {X : ℕ → Type*}
[∀ n, Nonempty (X n)] (W : ∀ n, ClassicalInputChannel (X n) (blockBasis I n)) :
Tendsto
(fun epsilon : ℝ => Filter.liminf
(fun n => optimizedTestingInformation (blockBasis I n) (W n) epsilon /
(((n + 1 : ℕ) : ℝ) : EReal)) atTop)
(nhdsWithin 0 (Set.Ioi 0)) (nhds (vanishingErrorCapacity W)) := by sorry
end QAlgorithms.Papers.Independent.Support50.P20
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be any nonempty type and let be a nonempty type for every natural number , including . Define , with its canonical Hilbert basis indexed by functions . Let assign a state on to every . Here a state is a bounded complex-linear operator that is self-adjoint with nonnegative real quadratic form, has , and admits a representation with . An effect means that both and are bounded self-adjoint operators with nonnegative real quadratic forms; write . For a probability mass function on and effects on , using real probability masses, set and . Define
where the infimum is in , , , and for positive finite . For each , a code consists of a positive integer , words , and effects on summing to the identity; its error is , and its words need not be distinct. Define
with the supremum and rate limit inferior taken in the extended reals. The declaration asserts that, as the real number tends to through strictly positive values,
in the extended-real topology. No finiteness or countability of or the , and no relation between the different functions , is assumed.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.