Definitions for The Converse Part of The Theorem for Quantum Hoeffding Bound
DefinitionQAlgorithms_Papers_Independent_Intakethird100_P04The objects the formalized statements of The Converse Part of The Theorem for Quantum Hoeffding Bound (arXiv:quant-ph/0611289) are stated in. Each was checked against the paper by a blind read-back and a moderator.
DensitySpectrum: PDF p.3, (12): complete eigenbasis of the given density, including its kernel. No countability, faithfulness or finite-dimensionality is imposed.spectralP: PDF p.3, (13): constructed Nussbaum--Szkola p weights.spectralQ: PDF p.3, (13): constructed Nussbaum--Szkola q weights. Completeness and trace normalization imply that each of p and q has total mass one.lowerMass: PDF p.4, (18),(21): p mass on the non-strict multiplicative event, including ties. Nonnegative sums retain infinity rather than defaulting to zero.upperMass: PDF p.4, (18),(22): q mass on the strict complementary event. No likelihood division or logarithm at a zero atom is used.iidOperator: PDF pp.1,4: (k+1)-copy operator in the complete tensor basis, realized as an ACTUAL bounded operator by the accepted boundedOperator. For bounded A, these coefficients always have a unique bounded realization A tensor ... tensor A, with norm at most norm(A)^(k+1). Thus reconstruction's off-domain zero branch cannot be reached. The coordinate Hilbert space is the completed tensor power.iidLaw: PDF p.4: (k+1) independent classical copies. For spectral probability weights these finite product laws also have mass one.logRate: PDF p.1, (1): natural log rate with positive denominator k+1. Zero probability is sent to minus infinity by ENNReal.log.hoeffdingB: PDF p.1, (1): supremum over all effect sequences and EXISTING extended alpha-rate limits, with the non-strict beta limsup constraint. No total limit selector or liminf replaces the limit. The zero-effect sequence is feasible for r>0 and gives exponent zero; unbounded exponents retain plus infinity.logMoment: PDF pp.2,7, (5),(31): extended natural log moment. Interior zero overlap, possible for orthogonal supports, has value minus infinity. Reuses imported traceAcceptance for the fractional-power product, with ordinary A^0=I. On states and 0<=s<=1 Schatten Holder ensures traceability and a trace in [0,1].
import Mathlib
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P20
import Definitions.Def_QAlgorithms_Papers_Independent_Frontier18_P18
/-!
# Nagaoka, quant-ph/0611289v1
Five nondeferred nodes on arbitrary complete complex Hilbert spaces.
Citations use PDF pages. The independent-paper override authorizes imports
from the pinned accepted library. Twenty other nodes, including the
endpoint-defective Theorem 1, are recorded in HARD.md and GAPS.json.
Tensor blocks use complete l2(I^(k+1)) coordinates of the completed tensor
power. B retains the EXISTING extended alpha-rate limit from (1).
Type-I errors and Petz traces use imported traceAcceptance directly;
no local alpha or petzOverlap definition is declared.
Compilation is pending the batch mechanical handoff.
-/
noncomputable section
open Filter
open scoped ENNReal NNReal BigOperators ComplexOrder
namespace QAlgorithms.Papers.Independent.Intakethird100.P04
attribute [local instance] Classical.propDecidable Classical.decEq
attribute [local instance] IsStarNormal.instContinuousFunctionalCalculus
open QAlgorithms.Papers.Independent.Dependency50.P13
(State operatorTrace relativeEntropy)
open QAlgorithms.Papers.Independent.Support50.P20
(Effect traceAcceptance IsHilbertCPTP BlockHilbert blockBasis)
open QAlgorithms.Papers.Independent.Frontier18.P18 (boundedOperator)
variable {H K : Type*}
variable [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H]
variable [NormedAddCommGroup K] [InnerProductSpace ℂ K] [CompleteSpace K]
/-- PDF p.3, (12): complete eigenbasis of the given density, including its
kernel. No countability, faithfulness or finite-dimensionality is imposed. -/
structure DensitySpectrum {I : Type*} {b : HilbertBasis I ℂ H}
(rho : State b) (J : Type*) where
basis : HilbertBasis J ℂ H
eigenvalue : J → ℝ≥0
eigenvector : ∀ j, rho.val.val (basis j) = (eigenvalue j : ℂ) • basis j
/-- PDF p.3, (13): constructed Nussbaum--Szkola p weights. -/
def spectralP {I J L : Type*} {b : HilbertBasis I ℂ H}
{rho sigma : State b} (x : DensitySpectrum rho J) (y : DensitySpectrum sigma L)
(w : J × L) : ℝ≥0∞ :=
(x.eigenvalue w.1 : ℝ≥0∞) *
ENNReal.ofReal (Complex.normSq (inner ℂ (x.basis w.1) (y.basis w.2)))
/-- PDF p.3, (13): constructed Nussbaum--Szkola q weights. Completeness and
trace normalization imply that each of p and q has total mass one. -/
def spectralQ {I J L : Type*} {b : HilbertBasis I ℂ H}
{rho sigma : State b} (x : DensitySpectrum rho J) (y : DensitySpectrum sigma L)
(w : J × L) : ℝ≥0∞ :=
(y.eigenvalue w.2 : ℝ≥0∞) *
ENNReal.ofReal (Complex.normSq (inner ℂ (x.basis w.1) (y.basis w.2)))
/-- PDF p.4, (18),(21): p mass on the non-strict multiplicative event,
including ties. Nonnegative sums retain infinity rather than defaulting to zero. -/
def lowerMass {Omega : Type*} (p q : Omega → ℝ≥0∞) (delta : ℝ≥0∞) : ℝ≥0∞ :=
∑' w, if p w ≤ delta * q w then p w else 0
/-- PDF p.4, (18),(22): q mass on the strict complementary event.
No likelihood division or logarithm at a zero atom is used. -/
def upperMass {Omega : Type*} (p q : Omega → ℝ≥0∞) (delta : ℝ≥0∞) : ℝ≥0∞ :=
∑' w, if delta * q w < p w then q w else 0
/-- PDF pp.1,4: (k+1)-copy operator in the complete tensor basis, realized
as an ACTUAL bounded operator by the accepted boundedOperator. For bounded A,
these coefficients always have a unique bounded realization A tensor ... tensor A,
with norm at most norm(A)^(k+1). Thus reconstruction's off-domain zero branch
cannot be reached. The coordinate Hilbert space is the completed tensor power. -/
def iidOperator {I : Type} (b : HilbertBasis I ℂ H)
(A : H →L[ℂ] H) (k : ℕ) : BlockHilbert I k →L[ℂ] BlockHilbert I k :=
boundedOperator (fun u v : Fin (k + 1) → I =>
∏ t, inner ℂ (b (u t)) (A (b (v t))))
/-- PDF p.4: (k+1) independent classical copies. For spectral probability
weights these finite product laws also have mass one. -/
def iidLaw {Omega : Type*} (p : Omega → ℝ≥0∞) (k : ℕ)
(w : Fin (k + 1) → Omega) : ℝ≥0∞ :=
∏ t, p (w t)
/-- PDF p.1, (1): natural log rate with positive denominator k+1.
Zero probability is sent to minus infinity by ENNReal.log. -/
def logRate (p : ℕ → ℝ≥0∞) (k : ℕ) : EReal :=
(((k + 1 : ℕ) : ℝ)⁻¹ : EReal) * ENNReal.log (p k)
/-- PDF p.1, (1): supremum over all effect sequences and EXISTING extended
alpha-rate limits, with the non-strict beta limsup constraint. No total limit
selector or liminf replaces the limit. The zero-effect sequence is feasible
for r>0 and gives exponent zero; unbounded exponents retain plus infinity. -/
def hoeffdingB {I : Type} (b : HilbertBasis I ℂ H)
(rho sigma : H →L[ℂ] H) (r : ℝ) : EReal :=
sSup {z | ∃ (T : (k : ℕ) → Effect (BlockHilbert I k)) (l : EReal),
Tendsto (logRate (fun k => ENNReal.ofReal
(1 - traceAcceptance (blockBasis I k) (T k).val (iidOperator b rho k)))) atTop (nhds l) ∧
Filter.limsup (logRate (fun k => ENNReal.ofReal
(traceAcceptance (blockBasis I k) (T k).val (iidOperator b sigma k)))) atTop ≤
-(r : EReal) ∧ z = -l}
/-- PDF pp.2,7, (5),(31): extended natural log moment. Interior zero overlap,
possible for orthogonal supports, has value minus infinity. Reuses imported
traceAcceptance for the fractional-power product, with ordinary A^0=I.
On states and 0<=s<=1 Schatten Holder ensures traceability and a trace in [0,1]. -/
def logMoment {I : Type*} (b : HilbertBasis I ℂ H)
(rho sigma : H →L[ℂ] H) (s : ℝ) : EReal :=
ENNReal.log (ENNReal.ofReal
(traceAcceptance b (CFC.rpow rho (1 - s)) (CFC.rpow sigma s)))
/-- Sanity: the zero test has type-I error one. -/
example {I : Type*} (b : HilbertBasis I ℂ H) (rho : H →L[ℂ] H) :
1 - traceAcceptance b 0 rho = 1 := by
simp [traceAcceptance, operatorTrace]
/-- Sanity: zero first law contributes zero mass at any threshold. -/
example {Omega : Type*} (q : Omega → ℝ≥0∞) (delta : ℝ≥0∞) :
lowerMass (fun _ => 0) q delta = 0 := by
simp [lowerMass]
/-- Sanity: the strict second event excludes ties. -/
example {Omega : Type*} (p : Omega → ℝ≥0∞) : upperMass p p 1 = 0 := by
simp [upperMass]
/-- Sanity: a product of unit coordinate weights remains unit. -/
example {Omega : Type*} (k : ℕ) (w : Fin (k + 1) → Omega) :
iidLaw (fun _ => 1) k w = 1 := by
simp [iidLaw]
/-- Sanity: the zero probability rate is minus infinity at one copy. -/
example : logRate (fun _ => 0) 0 = ⊥ := by
simp [logRate]
/-- Sanity: extended logarithms preserve zero overlap. -/
example : ENNReal.log (0 : ℝ≥0∞) = ⊥ := by simp
end QAlgorithms.Papers.Independent.Intakethird100.P04
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
QAlgorithms.Papers.Independent.Intakethird100.P04.DensitySpectrum
Let be a complete complex inner-product space, let be any index type, and let be a Hilbert basis of . Let be a state relative to : its underlying bounded operator satisfies , for every , and ; moreover, there exist sequences with and for every . For any index type , a DensitySpectrum of indexed by consists of a Hilbert basis of , a family of finite real numbers, and the identities
Zero eigenvalues are allowed, and no ordering of the eigenvalues is required. This structure specifies data and their required properties; it does not assert that such data exist for every state or index type.
QAlgorithms.Papers.Independent.Intakethird100.P04.hoeffdingB
Let be a complete complex inner-product space, any index type, a Hilbert basis, arbitrary bounded complex-linear operators, and . For each , put , with canonical basis , and define by selecting a bounded operator with coefficients , returning zero if none exists; define identically using . For bounded operators on , set , where this complex sum has value zero if its family is not summable. Write , regarded as an extended nonnegative real. Then the definition is
The limits and supremum are in the extended reals, and . The first rate must converge; the second is constrained only through its limsup. No positivity, trace-class, or trace-one assumptions are made on , and may be any real number. No admissible test sequence is asserted to exist; the supremum of an empty admissible set is . At , the block has one coordinate.
QAlgorithms.Papers.Independent.Intakethird100.P04.iidLaw
For any index type , any function , any , and any tuple , define
This is an extended nonnegative real number, with multiplication including . No normalization or finiteness condition is imposed on . The tuple has entries, so at the value is , rather than an empty product. If is empty, there is no tuple of this positive length.
QAlgorithms.Papers.Independent.Intakethird100.P04.iidOperator
Let be a complete complex inner-product space, let be any index type, let be a Hilbert basis of , let be any bounded complex-linear operator, and let . The result is a bounded operator on , where denotes the functions from to . Form the coefficient array
If there exists a bounded operator on this -space whose canonical-basis coefficients satisfy for all , the definition selects such a ; if no such bounded operator exists, it returns the zero operator. Here has coordinate one at and zero elsewhere. There are factors, so still gives one factor. No positivity, trace-class, normalization, countability, or nonemptiness assumption is imposed on or .
QAlgorithms.Papers.Independent.Intakethird100.P04.logMoment
Let be a complete complex inner-product space, let be any index type, let be a Hilbert basis, let be arbitrary bounded complex-linear operators, and let . Using the total functional-calculus real-power operation for bounded operators, define
where is regarded as an extended nonnegative real and the outer logarithm is natural, with . The complex sum is the total infinite-sum operation, which returns zero if the family is not summable. No positivity, self-adjointness, trace-class, normalization, invertibility, or restriction is assumed. In particular, a nonpositive real trace expression, or a nonsummable trace family, produces .
QAlgorithms.Papers.Independent.Intakethird100.P04.logRate
For every sequence and every , define the extended real number
where the logarithm is the natural extended logarithm, with and . The denominator is always positive, including at , so these infinite endpoint values remain respectively and . No probability, positivity away from zero, or finiteness assumption is imposed on .
QAlgorithms.Papers.Independent.Intakethird100.P04.lowerMass
For any index type , any functions , and any , define
an extended nonnegative sum. Equality belongs to the selected region. Neither nor is required to have total mass one, finite mass, or finite values, and may be empty or uncountable. The threshold may be zero or infinity; all products use extended nonnegative arithmetic, including . An empty selected region contributes zero.
QAlgorithms.Papers.Independent.Intakethird100.P04.spectralP
Let be a complete complex inner-product space with Hilbert basis , where are arbitrary index types. Let have underlying bounded operators , each self-adjoint with nonnegative quadratic forms, with , and each admitting a representation with . Suppose supplies a Hilbert basis and finite nonnegative eigenvalues satisfying , and supplies a Hilbert basis and finite nonnegative eigenvalues satisfying . For every , the definition assigns the extended nonnegative real number
The squared modulus is converted to an extended nonnegative real by replacing any negative real input by zero; here it is already nonnegative. Zero eigenvalues and zero overlaps are included. The definition itself makes no additional assertion about the sum of .
QAlgorithms.Papers.Independent.Intakethird100.P04.spectralQ
Let be a complete complex inner-product space with Hilbert basis , where are arbitrary index types. Let have underlying bounded operators , each self-adjoint with nonnegative quadratic forms, with , and each admitting a representation with . Suppose supplies a Hilbert basis and finite nonnegative eigenvalues satisfying , and supplies a Hilbert basis and finite nonnegative eigenvalues satisfying . For every , the definition assigns the extended nonnegative real number
The squared modulus is converted to an extended nonnegative real by replacing any negative real input by zero; here it is already nonnegative. Zero eigenvalues and zero overlaps are included. The definition itself makes no additional assertion about the sum of .
QAlgorithms.Papers.Independent.Intakethird100.P04.upperMass
For any index type , any functions , and any , define
an extended nonnegative sum. The selection inequality is strict, so points satisfying are excluded. Neither function is required to have total mass one, finite mass, or finite values, and may be empty or uncountable. The threshold may be zero or infinity; products use extended nonnegative arithmetic, including . An empty selected region contributes zero.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.