Local packet measurements update the residual common-phase law by Bayes rule
OpenQAlgorithms.Papers.Independent.Intakenext100.P33.bayesianPackets_updateLet , and start with . On each of the first packets perform a finite-outcome local quantum measurement, specified by bounded Kraus operators with a countable internal index and strong completeness . Fix an observed outcome string . Construct its effects from their convergent occupation sums. Write and define the actual Born likelihood and evidence by
Assume . Apply the tensor Kraus instrument to the first packets, leaving packets untouched, sum its internal branches in trace norm, and trace out the measured packets. This constructs an unnormalized residual ; the actual outcome probability is , and
All integral, Kraus-sum, effect and partial-trace constructions are genuine trace-class or bounded-operator realizations. The posterior law has mass one and is a normalized density. Each effect's Born weight equals its convergent sum of Kraus output norm squares. The residual consequently depends only on effects, regardless of their local Kraus realization.
This is finite-data quantum conditioning, with no exact finite-data phase identification or purity conclusion. The empty measurement string has likelihood one; means a scalar residual.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P33
noncomputable section
namespace QAlgorithms.Papers.Independent.Intakenext100.P33
open Filter MeasureTheory
open scoped Topology Matrix ComplexConjugate
attribute [local instance] Classical.propDecidable
open QAlgorithms.Papers.Independent.Dependency50.P39
-- Accepted scalar phase; positive exponential, differing only by commuting scalars.
open QAlgorithms.Papers.Independent.Information100.P75 (phase)
/-- arXiv:quant-ph/0111157v1, PDF p.5, Eqs.(22)–(23).
Actual local quantum instruments on the first K packets give the computed
likelihood, evidence and posterior on M untouched packets. Arbitrary countable
internal Kraus realizations are allowed; after tracing the measured packets,
the answer depends only on the selected effects. Evidence must be positive.
K=0 and M=0 respectively mean no data and the scalar residual density. -/
theorem bayesianPackets_update {K M O : ℕ} (a : ℝ) (ha : 0 ≤ a)
(A : Fin K → FockLocalInstrument O) (d : Fin K → Fin O)
(hZ : 0 < packetEvidence A d a) :
(∀ i : Fin K,
HasFockMatrix (localEffect (A i) (d i)) (fun m n => ∑' k : ℕ × ℕ,
star ((A i).kraus (d i) k.1 (fockKet m) k.2) *
(A i).kraus (d i) k.1 (fockKet n) k.2) ∧
(∀ theta, Summable (fun r : ℕ =>
‖(A i).kraus (d i) r (coherentKet ((a : ℂ) * phase theta))‖ ^ 2) ∧
(inner (𝕜 := ℂ) (coherentKet ((a : ℂ) * phase theta))
(localEffect (A i) (d i) (coherentKet ((a : ℂ) * phase theta)))).re =
∑' r : ℕ, ‖(A i).kraus (d i) r (coherentKet ((a : ℂ) * phase theta))‖ ^ 2)) ∧
(∀ r, HasPacketMatrix (eventKraus M A d r) (fun m n =>
if (m ∘ Fin.natAdd K) = (n ∘ Fin.natAdd K) then
∏ i : Fin K, (A i).kraus (d i) (r i)
(fockKet (n (Fin.castAdd M i))) (m (Fin.castAdd M i)) else 0)) ∧
TraceHasSum (eventTerm M A d a) (eventOutput M A d a) ∧
(Summable (fun q => eventOutput M A d a (packetBasis q) q) ∧
(∑' q, eventOutput M A d a (packetBasis q) q).re = packetEvidence A d a) ∧
(∀ m n, Summable (fun q : Fin K → ℕ =>
eventOutput M A d a (packetBasis (Fin.append q n)) (Fin.append q m))) ∧
HasPacketMatrix (packetResidual M A d a) (fun m n => ∑' q : Fin K → ℕ,
eventOutput M A d a (packetBasis (Fin.append q n)) (Fin.append q m)) ∧
TraceIntegral phaseLaw (fun theta => (packetLikelihood A d a theta : ℂ) •
packetProjector (fun _ : Fin M => (a : ℂ) * phase theta))
(packetResidual M A d a) ∧
IsProbabilityMeasure (posteriorPhaseLaw A d a) ∧
IsPacketDensity (((packetEvidence A d a)⁻¹ : ℂ) • packetResidual M A d a) ∧
TraceIntegral (posteriorPhaseLaw A d a)
(fun theta => packetProjector (fun _ : Fin M => (a : ℂ) * phase theta))
(((packetEvidence A d a)⁻¹ : ℂ) • packetResidual M A d a) := by sorry
end QAlgorithms.Papers.Independent.Intakenext100.P33
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
Let , let be real, and work on , with basis , and , with basis . For each , suppose an instrument consists of bounded operators on , indexed by and , satisfying, for every ,
as a sum of vectors; fix outcomes . In what follows, selecting an operator from a prescribed matrix means choosing a bounded operator having that matrix if one exists, and returning zero otherwise. Let be a selected vector with coefficients , or zero if no such vector exists; let have the product of these coefficients over its modes when square summable, and zero otherwise; write , where ; and let . Define by selecting the operator with matrix
and define
Assume . For , let be selected from the matrix
on ; put , let be the totalized sum in the bounded-operator space, and select from the matrix
Call an operator nuclear when it has an operator-norm-convergent expansion with , and let be the real infimum of these sums, with value zero if there are no expansions. Write to mean that is Bochner integrable with integral , that is nuclear and is nuclear almost everywhere, and that there exist finite-range Bochner-integrable for which and are nuclear almost everywhere, is integrable, and . Then every has exactly its displayed matrix, and for every and , the squared norms form a summable family with sum . Every has exactly its displayed matrix. The operator , every finite partial sum , and every difference are nuclear, and along all finite subsets . The complex diagonal of is summable and
For every , the defining series for is summable, and has exactly that matrix. Furthermore,
holds. The measure with density relative to , where truncates negative values to zero, is a probability measure; is nuclear and self-adjoint, satisfies for every , and has summable complex diagonal with sum one; and
holds. The quantifiers include and , with empty products equal to one and a singleton; if and , no outcome function of the required type exists.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.