Proposition 6.5
OpenQAlgorithms.Papers.Independent.Support50.P12.proposition_6_5Let and be concrete unital algebras on nonzero finite-dimensional complex Hilbert spaces, and let be complex linear. The adjoint is for . A substochastic map has a Schwarz-contracting adjoint; stochastic additionally means trace preserving. Let . Powers are zero on kernels and traces are not normalized. Write for the unnormalized Petz Rényi entropy, , , and . Logarithms are natural and . The Petz map is .
Let and suppose admits two nonzero orthogonal projections. There is no operator convex admissible with and strictly increasing on the attained extended range of such that
Thus even monotone reparametrization of such a rich divergence cannot represent this distance.
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P12
noncomputable section
namespace QAlgorithms.Papers.Independent.Support50.P12
open Filter MeasureTheory Set
open QAlgorithms.Papers.Independent.Information100.P60 (Mat copies)
open QAlgorithms.Papers.Independent.Information100.P30 (MatrixMap)
open QAlgorithms.Papers.Independent.Dependency50.P28 (tr LE logNonnegativeExtended pinch)
open QAlgorithms.Papers.Independent.Dependency50.P21 (channelAdjoint spectralFunction petz relativeEntropy)
open QAlgorithms.Papers.Independent.Frontier18.P17 (hs)
open QAlgorithms.Papers.Independent.Next46.P23 (Supports)
open QAlgorithms.Papers.Independent.Support21.P03 (ExtendedConvex)
open QAlgorithms (spectralProjector)
open QAlgorithms.Learning (IsDistribution)
open QAlgorithms.Papers.Independent.Information100.P04 (absOp)
open scoped BigOperators Matrix ComplexConjugate ComplexOrder ENNReal Topology Kronecker
attribute [local instance] Classical.propDecidable Classical.decEq
open QAlgorithms.Papers.Independent.Information100.P71 (supportPower)
variable {I J K : Type} [Fintype I] [DecidableEq I] [Fintype J] [DecidableEq J]
[Fintype K] [DecidableEq K]
/-- Proposition 6.5, PDF p.30: no strictly increasing function on the attained
extended divergence range represents a positive-rate Hoeffding distance. -/
theorem proposition_6_5 [Nonempty I] (A : Alg I) (ha : TwoProjections A) (r : ℝ) (hr : 0 < r) :
¬ ∃ f, DivergenceGenerator f ∧ SemidefiniteOperatorConvex f ∧ Rich f 6 ∧
MonotoneRepresentation A (fun X Y => hoeffding X Y r) f := by sorry
end QAlgorithms.Papers.Independent.Support50.P12
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let be a finite, nonempty index type with decidable equality, let be a unital complex star subalgebra of , and assume that contains nonzero Hermitian idempotents with . Fix a real . For positive semidefinite , set
where spectral powers take value at zero for every exponent and for , otherwise. For a real function , define to be
when , and otherwise its totalized extended-real limit after replacing by as , with . There does not exist an satisfying all of the following conditions: is continuous on and has an extended-real limit at infinity; for every natural matrix size , positive semidefinite , and , spectral calculus satisfies ; the selected representing measure described below has support of cardinality at least six; and there is a strictly increasing function , defined on exactly the extended-real values attained by positive semidefinite trace-one , such that for every such pair. The representing measure is a selected measure on for the function equal to away from zero and to its totalized right-limit value at zero: it is selected to satisfy, for some real with ,
with integrable and each displayed kernel integrable; if no measure satisfies this predicate, the selection is unspecified. The richness condition concerns this selected measure.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.