Example (new estimates) 3.6
OpenQAlgorithms.Papers.Independent.Support50.P12.example_3_6Let have dimension and fix an orthonormal basis for transpose. For real , define , , and . The Schwarz norm is the least nonnegative with for all , and is infinite if no such exists.
For , is Schwarz and
Its defect has the six-term expansion in the source, and for , , it is bounded below by . Matrix units give the diagonal defect . The Schwarz norm tends to infinity as , whereas the operator norm is one for ; the norms are strictly different when . Some are Schwarz contractions without being two-positive. Further, is Schwarz exactly for , with Schwarz norm tending to infinity at that upper endpoint. For , the -th diagonal defect of at equals for every real , so is Schwarz exactly when . These are the authors' new estimates and conclusions; classification ranges attributed to Tomiyama remain background.
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]
/-- Example 3.6 (new estimates), PDF pp.12–13. The classification ranges expressly
attributed to Tomiyama are background; this statement retains the new calculations. -/
theorem example_3_6 (hd : 1 < Fintype.card I) :
let d : ℝ := Fintype.card I
(∃ e : ℝ, SchwarzBound ⊤ (transposeNoise (I := I) e) 1 ∧
¬ AlgebraKPositive ⊤ (transposeNoise (I := I) e) 2) ∧
(∀ e ∈ Ico (0 : ℝ) 1, ∀ c ≥ (0 : ℝ), ∀ X : Mat I,
(c : ℂ) • transposeNoise e (Xᴴ * X) - transposeNoise e (Xᴴ) * transposeNoise e X =
transposeDefect e c X ∧
LE (((c * e - d * (1 - e) ^ 2 - 2 * e * (1 - e) * Real.sqrt d - e ^ 2) / d : ℂ) •
(tr (Xᴴ * X) : ℂ) • (1 : Mat I))
((c : ℂ) • transposeNoise e (Xᴴ * X) -
transposeNoise e (Xᴴ) * transposeNoise e X)) ∧
(∀ e ∈ Ioo (0 : ℝ) 1, AlgebraSchwarzMap ⊤ (transposeNoise (I := I) e) ∧
ENNReal.ofReal (d * (1 - e) ^ 2 / e) ≤ snorm ⊤ (transposeNoise (I := I) e) ∧
snorm ⊤ (transposeNoise (I := I) e) ≤ ENNReal.ofReal
((d * (1 - e) ^ 2 + 2 * e * (1 - e) * Real.sqrt d + e ^ 2) / e)) ∧
(∀ e ∈ Ioo (0 : ℝ) 1, ∀ i j : I, i ≠ j →
let X : Mat I := Matrix.single i j 1
let S := (((snorm ⊤ (transposeNoise (I := I) e)).toReal : ℂ) • transposeNoise e (Xᴴ * X) -
transposeNoise e (Xᴴ) * transposeNoise e X) i i
0 ≤ S.re ∧ S.re = (snorm ⊤ (transposeNoise (I := I) e)).toReal * e / d - (1 - e) ^ 2) ∧
Tendsto (fun e : ℝ => snorm ⊤ (transposeNoise (I := I) e)) (𝓝[>] 0) (𝓝 ⊤) ∧
(∀ e ∈ Icc (0 : ℝ) (1 + 1 / (d - 1)), pnorm (transposeNoise (I := I) e) = 1) ∧
(∀ e ∈ Ioo (0 : ℝ) 1, d < (1 - e) ^ 2 / e →
ENNReal.ofReal (pnorm (transposeNoise (I := I) e)) < snorm ⊤ (transposeNoise (I := I) e)) ∧
(∀ e : ℝ, AlgebraSchwarzMap ⊤ (identityNoise (I := I) e) ↔ 0 ≤ e ∧ e < 1 + 1 / (d - 1)) ∧
Tendsto (fun e : ℝ => snorm ⊤ (identityNoise (I := I) e))
(𝓝[<] (1 + 1 / (d - 1))) (𝓝 ⊤) ∧
(∀ e c : ℝ, ∀ i j : I, i ≠ j →
let X : Mat I := Matrix.single i j 1
let D : Mat I := (c : ℂ) • transposeMix e (Xᴴ * X) -
transposeMix e (Xᴴ) * transposeMix e X
(D i i).re = -(1 - e) ^ 2) ∧
(∀ e : ℝ, AlgebraSchwarzMap ⊤ (transposeMix (I := I) e) ↔ e = 1) := by sorry
end QAlgorithms.Papers.Independent.Support50.P12
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For every finite index type with decidable equality and , define , , and . Write when is positive semidefinite, and . For a matrix map , let consist of the real such that sends every positive semidefinite matrix to a positive semidefinite matrix and
put , where , and
The assertion is the conjunction of all the following statements: there exists a real with for which applying to each block of some positive semidefinite block matrix fails to preserve positive semidefiniteness; for every , every , and every , the matrix equals
and satisfies
for every , is nonempty and
for every such and every distinct , setting , the matrix unit with entry at , and , one has ; as from above; for every ; whenever and , ; for every real , is nonempty exactly when ; as approaches from below; for every real and every distinct , setting gives
and, for every real , is nonempty exactly when . The notation denotes conversion from to , which assigns the value .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.