LIDT definitions: LDT / SelfImprovement / MatrixRealization / Canonical / StrongDuality / Basic
DefinitionLIDT_MIPStarRE_LDT_SelfImprovement_MatrixRealization_Canonical_StrongDuality_Basiclidt
This bundle develops the analytic foundations for strong duality of the canonical matrix SDP. It proves feasibility and trace identities, controls feasible matrix norms, and packages the relevant diagonal-block, constraint, and objective maps as continuous real-linear maps.
Principal definitions:
matrixSdpCanonicalPrimalFeasible_nonempty: exhibits a feasible canonical primal point.matrixSdpCanonicalConstraintOperator_trace_eq: identifies the trace of the constraint image.matrixSdpCanonicalNonnegative_norm_le_constraint_trace_re: controls a PSD variable by the real constraint trace.matrixSdpCanonicalPrimalFeasible_trace_eq: gives the trace of a feasible primal matrix.matrixSdpCanonicalPrimalFeasible_norm_le: gives a uniform norm bound for feasible matrices.matrixSdpCanonicalDiagonalBlockCLM: the continuous linear map extracting a diagonal block.matrixSdpCanonicalConstraintOperatorCLM: the continuous constraint map.matrixSdpCanonicalPrimalObjectiveCLM: the continuous primal objective functional.
Formalization credit: Sirui Lu, Ruixuan Deng, Yanqiao Zhu, and Zhengfeng Ji (arXiv:2609.19814). Mathematical source: Zhengfeng Ji, Anand Natarajan, Thomas Vidick, John Wright, and Henry Yuen (arXiv:2009.12982).
Definition code
-- Generated from MIPStarRE/LDT/SelfImprovement/MatrixRealization/Canonical/StrongDuality/Basic.lean by skeleton subtraction.
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_AxisParallelLine
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DiagonalLine
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_Distribution
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DistributionAvg
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_DistributionUniformSums
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_LinePolynomials
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_LowDegreePolynomial
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_MeasurementLift
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_OpFamily
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_OperatorExpectations
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_PMFAverages
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_PMFUniformAverages
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_ParametersBase
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_ParametersFiniteAnswers
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_QuantumState
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_SubMeasurementCore
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_SubMeasurementFamilies
import Definitions.Def_LIDT_MIPStarRE_LDT_Basic_TensorPlacement
import Definitions.Def_LIDT_MIPStarRE_LDT_ExpansionHypercubeGraph_Defs_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_ExpansionHypercubeGraph_Defs_Fourier
import Definitions.Def_LIDT_MIPStarRE_LDT_ExpansionHypercubeGraph_MatrixRealization_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_ExpansionHypercubeGraph_MatrixRealization_TraceForms
import Definitions.Def_LIDT_MIPStarRE_LDT_GlobalVariance_Defs_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_GlobalVariance_Defs_Families
import Definitions.Def_LIDT_MIPStarRE_LDT_GlobalVariance_Defs_Operators
import Definitions.Def_LIDT_MIPStarRE_LDT_MainInductionStep_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_NaimarkCore
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Projectivization
import Definitions.Def_LIDT_MIPStarRE_LDT_MakingMeasurementsProjective_Statements
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Completion
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_BipartiteSelfConsistency_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_CauchySchwarz
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_ComparisonCore
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_DistanceBounds
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichMain_Completeness
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_ApproxDelta
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_Core
import Definitions.Def_LIDT_MIPStarRE_LDT_Preliminaries_SwitchSandwichPrep_InnerProduct
import Definitions.Def_LIDT_MIPStarRE_LDT_SelfImprovement_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_SelfImprovement_MatrixRealization_Base
import Definitions.Def_LIDT_MIPStarRE_LDT_SelfImprovement_MatrixRealization_Canonical
import Definitions.Def_LIDT_MIPStarRE_LDT_SelfImprovement_MatrixRealization_CanonicalPrimal
import Definitions.Def_LIDT_MIPStarRE_LDT_Tactic_QuantumNonneg
import Definitions.Def_LIDT_MIPStarRE_LDT_Test_Defs
import Definitions.Def_LIDT_MIPStarRE_LDT_Test_StrategyCore
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_Basic
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_BlockDiagonal
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_NormalizedTrace
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_Order
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_TracePairing
import Definitions.Def_LIDT_MIPStarRE_Quantum_Measurement
import Mathlib
import Mathlib.Analysis.Fourier.ZMod
import Mathlib.Analysis.MeanInequalitiesPow
import Mathlib.Probability.ProbabilityMassFunction.Integrals
import Mathlib.Probability.ProbabilityMassFunction.Monad
namespace LIDT
namespace MIPStarRE.LDT.Preliminaries
end MIPStarRE.LDT.Preliminaries
/-!
# Section 9 -- Canonical SDP strong-duality preliminaries
This module contains the feasibility, compactness, closedness, and
objective-continuity lemmas used in the finite-dimensional strong-duality
argument for the canonical matrix SDP.
## References
- `references/ldt-paper/self_improvement.tex`
-/
@[expose] public section
namespace MIPStarRE.LDT.SelfImprovement
open MIPStarRE.LDT
open MIPStarRE.LDT.ExpansionHypercubeGraph
open MIPStarRE.LDT.GlobalVariance
open MIPStarRE.LDT.MakingMeasurementsProjective
open Filter
open scoped BigOperators MatrixOrder _root_.Matrix ComplexOrder _root_.Matrix.Norms.Elementwise Topology
/-- The canonical primal SDP has a feasible point, supplied by the explicit
strict primal submeasurement. -/
theorem matrixSdpCanonicalPrimalFeasible_nonempty
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
∃ X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model),
MatrixSdpCanonicalPrimalFeasible params model X :=
⟨matrixSdpCanonicalPrimalBlockMatrix params model
(matrixSdpStrictPrimalSubmeasurement params model),
matrixSdpCanonicalStrictPrimalBlockMatrix_feasible params model⟩
/-- The canonical equality-constraint operator preserves trace. -/
theorem matrixSdpCanonicalConstraintOperator_trace_eq
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model)) :
Matrix.trace (matrixSdpCanonicalConstraintOperator params model X) =
Matrix.trace X := by
rw [matrixSdpCanonical_trace_eq_sum_diagonalBlock params model X]
unfold matrixSdpCanonicalConstraintOperator
rw [Matrix.trace_sum]
/-- A PSD canonical primal variable is norm-controlled by the real trace of its
constraint image. -/
theorem matrixSdpCanonicalNonnegative_norm_le_constraint_trace_re
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model))
(hX : 0 ≤ X) :
‖X‖ ≤ Complex.re
(Matrix.trace (matrixSdpCanonicalConstraintOperator params model X)) := by
rw [matrixSdpCanonicalConstraintOperator_trace_eq params model X]
exact MIPStarRE.Quantum.norm_le_trace_re_of_nonneg hX
/-- A feasible canonical primal matrix has trace equal to the base Hilbert-space dimension. -/
theorem matrixSdpCanonicalPrimalFeasible_trace_eq
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model))
(hX : MatrixSdpCanonicalPrimalFeasible params model X) :
Matrix.trace X = (Fintype.card model.space.carrier : ℂ) := by
calc
Matrix.trace X = Matrix.trace (matrixSdpCanonicalConstraintOperator params model X) := by
exact (matrixSdpCanonicalConstraintOperator_trace_eq params model X).symm
_ = Matrix.trace (1 : MatrixOperator model.space) := by rw [hX.constraintEqOne]
_ = (Fintype.card model.space.carrier : ℂ) := by rw [Matrix.trace_one]
/-- Feasible canonical primal matrices have uniformly bounded elementwise norm. -/
theorem matrixSdpCanonicalPrimalFeasible_norm_le
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model))
(hX : MatrixSdpCanonicalPrimalFeasible params model X) :
‖X‖ ≤ (Fintype.card model.space.carrier : ℝ) := by
have hnorm := MIPStarRE.Quantum.norm_le_trace_re_of_nonneg hX.nonnegative
rw [matrixSdpCanonicalPrimalFeasible_trace_eq params model X hX] at hnorm
simpa using hnorm
/-- The feasible set of the canonical primal SDP is bounded in the elementwise matrix norm. -/
theorem matrixSdpCanonicalPrimalFeasible_isBounded
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
Bornology.IsBounded
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) |
MatrixSdpCanonicalPrimalFeasible params model X} := by
rw [isBounded_iff_forall_norm_le]
exact ⟨(Fintype.card model.space.carrier : ℝ), fun X hX =>
matrixSdpCanonicalPrimalFeasible_norm_le params model X hX⟩
/-- The projection onto one diagonal block of the canonical primal matrix, as a
continuous real-linear map. -/
noncomputable def matrixSdpCanonicalDiagonalBlockCLM
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(b : MatrixSdpCanonicalBlockIndex params) :
MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) →L[ℝ]
MatrixOperator model.space :=
ContinuousLinearMap.mk
((matrixSdpCanonicalDiagonalBlockLinearMap params model b).restrictScalars ℝ)
(continuous_matrix fun i j => continuous_apply_apply (b, i) (b, j))
@[simp]
theorem matrixSdpCanonicalDiagonalBlockCLM_apply
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(b : MatrixSdpCanonicalBlockIndex params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model)) :
matrixSdpCanonicalDiagonalBlockCLM params model b X =
matrixSdpCanonicalDiagonalBlock params model X b :=
rfl
/-- The canonical equality-constraint operator as a continuous real-linear map. -/
noncomputable def matrixSdpCanonicalConstraintOperatorCLM
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) →L[ℝ]
MatrixOperator model.space :=
∑ b : MatrixSdpCanonicalBlockIndex params,
matrixSdpCanonicalDiagonalBlockCLM params model b
@[simp]
theorem matrixSdpCanonicalConstraintOperatorCLM_apply
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model)) :
matrixSdpCanonicalConstraintOperatorCLM params model X =
matrixSdpCanonicalConstraintOperator params model X := by
simp [matrixSdpCanonicalConstraintOperatorCLM, matrixSdpCanonicalConstraintOperator]
/-- The canonical primal objective as a continuous real-linear functional. -/
noncomputable def matrixSdpCanonicalPrimalObjectiveCLM
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) →L[ℝ] ℝ :=
MIPStarRE.Quantum.realTracePairingCLM
(matrixSdpCanonicalObjectiveOperator params model)
/-- The canonical primal equality-constraint operator is continuous. -/
theorem continuous_matrixSdpCanonicalConstraintOperator
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
Continuous fun X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) =>
matrixSdpCanonicalConstraintOperator params model X := by
have h :
(fun X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) =>
matrixSdpCanonicalConstraintOperator params model X) =
fun X => matrixSdpCanonicalConstraintOperatorCLM params model X := by
funext X
exact (matrixSdpCanonicalConstraintOperatorCLM_apply params model X).symm
rw [h]
exact (matrixSdpCanonicalConstraintOperatorCLM params model).continuous
/-- A canonical block-diagonal operator is Hermitian when all diagonal blocks are
Hermitian. -/
theorem matrixSdpCanonicalBlockDiagonal_isHermitian
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(B : MatrixSdpCanonicalBlockIndex params → MatrixOperator model.space)
(hB : ∀ b, (B b).IsHermitian) :
(matrixSdpCanonicalBlockDiagonal params model B).IsHermitian := by
rw [matrixSdpCanonicalBlockDiagonal_eq_reindex_blockDiagonal]
exact (Matrix.isHermitian_blockDiagonal_iff.mpr hB).reindex
(Equiv.prodComm model.space.carrier (MatrixSdpCanonicalBlockIndex params))
/-- The canonical dual block operator is Hermitian when the paper dual matrix is
Hermitian. -/
theorem matrixSdpCanonicalDualOperator_isHermitian
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(Z : MatrixOperator model.space)
(hZ : Z.IsHermitian) :
(matrixSdpCanonicalDualOperator params model Z).IsHermitian := by
rw [matrixSdpCanonicalDualOperator]
exact matrixSdpCanonicalBlockDiagonal_isHermitian params model
(matrixSdpCanonicalDualOperatorBlockFamily params model Z) fun _ => hZ
/-- The canonical objective block operator is Hermitian. -/
theorem matrixSdpCanonicalObjectiveOperator_isHermitian
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
(matrixSdpCanonicalObjectiveOperator params model).IsHermitian := by
rw [matrixSdpCanonicalObjectiveOperator]
refine matrixSdpCanonicalBlockDiagonal_isHermitian params model
(matrixSdpCanonicalObjectiveBlockFamily params model) ?_
intro b
cases b with
| none => simp [matrixSdpCanonicalObjectiveBlockFamily]
| some g =>
exact (Matrix.nonneg_iff_posSemidef.mp
(matrixAveragedPointOperator_nonneg params model g)).isHermitian
/-- The canonical dual slack block operator is Hermitian when the paper dual
matrix is Hermitian. -/
theorem matrixSdpCanonicalDualOperator_sub_objectiveOperator_isHermitian
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(Z : MatrixOperator model.space)
(hZ : Z.IsHermitian) :
(matrixSdpCanonicalDualOperator params model Z -
matrixSdpCanonicalObjectiveOperator params model).IsHermitian :=
(matrixSdpCanonicalDualOperator_isHermitian params model Z hZ).sub
(matrixSdpCanonicalObjectiveOperator_isHermitian params model)
/-- The canonical dual block trace pairing equals the paper dual trace pairing
against the canonical constraint image. -/
theorem matrixSdpCanonicalDualOperator_trace_constraint
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model))
(Z : MatrixOperator model.space) :
Complex.re (Matrix.trace
(matrixSdpCanonicalDualOperator params model Z * X)) =
Complex.re (Matrix.trace
(Z * matrixSdpCanonicalConstraintOperator params model X)) := by
congr 1
rw [matrixSdpCanonicalDualOperator]
rw [matrixSdpCanonicalBlockDiagonal_trace_mul_left]
simp only [matrixSdpCanonicalDualOperatorBlockFamily_apply]
rw [← Matrix.trace_sum]
rw [← Finset.mul_sum]
rfl
/-- The canonical equality-constraint image of a positive canonical primal
matrix is positive semidefinite. -/
theorem matrixSdpCanonicalConstraintOperator_nonneg_of_nonnegative
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model)}
(hX : 0 ≤ X) :
0 ≤ matrixSdpCanonicalConstraintOperator params model X := by
unfold matrixSdpCanonicalConstraintOperator
exact Finset.sum_nonneg fun b _ =>
matrixSdpCanonicalDiagonalBlock_nonneg params model hX b
/-- The canonical equality-constraint image of a positive canonical primal
matrix is Hermitian. -/
theorem matrixSdpCanonicalConstraintOperator_isHermitian_of_nonnegative
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model)}
(hX : 0 ≤ X) :
(matrixSdpCanonicalConstraintOperator params model X).IsHermitian :=
(Matrix.nonneg_iff_posSemidef.mp
(matrixSdpCanonicalConstraintOperator_nonneg_of_nonnegative params model hX)).isHermitian
/-- A trace-pairing separator against every positive semidefinite canonical primal
matrix can be converted into the paper-form dual feasibility inequalities.
This is a separator-conversion lemma for the later zero-gap argument, not the
strong-duality theorem itself. -/
theorem matrixSdpCanonicalDualConstraint_nonneg_of_trace_pairing_nonneg
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params)
(Z : MatrixOperator model.space)
(hHerm :
(matrixSdpCanonicalDualOperator params model Z -
matrixSdpCanonicalObjectiveOperator params model).IsHermitian)
(htrace :
∀ X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model),
0 ≤ X →
0 ≤ Complex.re (Matrix.trace
((matrixSdpCanonicalDualOperator params model Z -
matrixSdpCanonicalObjectiveOperator params model) * X))) :
∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g := by
have hcanonical :
0 ≤ matrixSdpCanonicalDualOperator params model Z -
matrixSdpCanonicalObjectiveOperator params model :=
MIPStarRE.Quantum.nonneg_of_trace_mul_nonneg_of_isHermitian hHerm htrace
exact matrixSdpDualFeasible_of_canonicalDualConstraint_nonneg params model Z hcanonical
/-- The feasible set of the canonical primal SDP is closed. -/
theorem matrixSdpCanonicalPrimalFeasible_isClosed
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
IsClosed
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) |
MatrixSdpCanonicalPrimalFeasible params model X} := by
classical
have hnonneg : IsClosed
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) | 0 ≤ X} :=
MIPStarRE.Quantum.isClosed_op_nonnegative
(ι := (matrixSdpCanonicalBlockHilbertSpace params model).carrier)
have hconstraint : IsClosed
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) |
matrixSdpCanonicalConstraintOperator params model X = (1 : MatrixOperator model.space)} :=
isClosed_eq (continuous_matrixSdpCanonicalConstraintOperator params model) continuous_const
rw [show
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) |
MatrixSdpCanonicalPrimalFeasible params model X} =
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) | 0 ≤ X} ∩
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) |
matrixSdpCanonicalConstraintOperator params model X =
(1 : MatrixOperator model.space)} by
ext X
constructor
· intro hX
exact ⟨hX.nonnegative, hX.constraintEqOne⟩
· intro hX
exact ⟨hX.1, hX.2⟩]
exact hnonneg.inter hconstraint
/-- The paper-form canonical dual feasible set is closed. -/
theorem matrixSdpCanonicalDualFeasible_isClosed
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
IsClosed
{Z : MatrixOperator model.space |
∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g} := by
classical
rw [show
{Z : MatrixOperator model.space |
∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g} =
⋂ g : Polynomial params,
{Z : MatrixOperator model.space |
0 ≤ matrixSdpDualSlackOperator params model Z g} by
ext Z
simp]
refine isClosed_iInter fun g => ?_
have hslack : Continuous fun Z : MatrixOperator model.space =>
matrixSdpDualSlackOperator params model Z g := by
unfold matrixSdpDualSlackOperator
exact continuous_id.sub continuous_const
simpa [Set.preimage] using
(MIPStarRE.Quantum.isClosed_op_nonnegative (ι := model.space.carrier)).preimage hslack
/-- The paper-form canonical dual objective is continuous. -/
theorem continuous_matrixSdpDualObjective
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
Continuous fun Z : MatrixOperator model.space => matrixSdpDualObjective model Z := by
unfold matrixSdpDualObjective
exact Complex.continuous_re.comp continuous_id.matrix_trace
/-- The strict-witness-bounded dual feasible sublevel is closed. -/
theorem matrixSdpCanonicalDualFeasibleSublevel_isClosed
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
IsClosed
{Z : MatrixOperator model.space |
(∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g) ∧
matrixSdpDualObjective model Z ≤
matrixSdpDualObjective model (matrixSdpStrictDualWitness model)} := by
have hfeasible := matrixSdpCanonicalDualFeasible_isClosed params model
have hsublevel : IsClosed
{Z : MatrixOperator model.space |
matrixSdpDualObjective model Z ≤
matrixSdpDualObjective model (matrixSdpStrictDualWitness model)} :=
isClosed_le (continuous_matrixSdpDualObjective params model) continuous_const
rw [show
{Z : MatrixOperator model.space |
(∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g) ∧
matrixSdpDualObjective model Z ≤
matrixSdpDualObjective model (matrixSdpStrictDualWitness model)} =
{Z : MatrixOperator model.space |
∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g} ∩
{Z : MatrixOperator model.space |
matrixSdpDualObjective model Z ≤
matrixSdpDualObjective model (matrixSdpStrictDualWitness model)} by
ext Z
rfl]
exact hfeasible.inter hsublevel
/-- The strict-witness-bounded dual feasible sublevel is norm-bounded. -/
theorem matrixSdpCanonicalDualFeasibleSublevel_isBounded
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
Bornology.IsBounded
{Z : MatrixOperator model.space |
(∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g) ∧
matrixSdpDualObjective model Z ≤
matrixSdpDualObjective model (matrixSdpStrictDualWitness model)} := by
rw [isBounded_iff_forall_norm_le]
refine ⟨matrixSdpDualObjective model (matrixSdpStrictDualWitness model), fun Z hZ => ?_⟩
calc
‖Z‖ ≤ Complex.re (Matrix.trace Z) :=
MIPStarRE.Quantum.norm_le_trace_re_of_nonneg
(matrixSdpDualPositive_of_dualFeasible params model Z hZ.1)
_ = matrixSdpDualObjective model Z := by rfl
_ ≤ matrixSdpDualObjective model (matrixSdpStrictDualWitness model) := hZ.2
/-- The strict-witness-bounded dual feasible sublevel is compact. -/
theorem matrixSdpCanonicalDualFeasibleSublevel_isCompact
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
IsCompact
{Z : MatrixOperator model.space |
(∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g) ∧
matrixSdpDualObjective model Z ≤
matrixSdpDualObjective model (matrixSdpStrictDualWitness model)} :=
Metric.isCompact_of_isClosed_isBounded
(matrixSdpCanonicalDualFeasibleSublevel_isClosed params model)
(matrixSdpCanonicalDualFeasibleSublevel_isBounded params model)
/-- The paper-form canonical dual objective attains its minimum on the feasible set. -/
theorem matrixSdpCanonicalDualObjective_exists_isMinOn
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
∃ Z : MatrixOperator model.space,
(∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g) ∧
∀ W : MatrixOperator model.space,
(∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model W g) →
matrixSdpDualObjective model Z ≤ matrixSdpDualObjective model W := by
let c := matrixSdpDualObjective model (matrixSdpStrictDualWitness model)
let S : Set (MatrixOperator model.space) :=
{Z | (∀ g : Polynomial params,
0 ≤ matrixSdpDualSlackOperator params model Z g) ∧
matrixSdpDualObjective model Z ≤ c}
have hScompact : IsCompact S := by
simpa [S, c] using matrixSdpCanonicalDualFeasibleSublevel_isCompact params model
have hSne : S.Nonempty := by
refine ⟨matrixSdpStrictDualWitness model, ?_⟩
exact ⟨matrixSdpStrictDualWitness_dualFeasible params model, le_rfl⟩
obtain ⟨Z, hZS, hZmin⟩ := hScompact.exists_isMinOn hSne
(continuous_matrixSdpDualObjective params model).continuousOn
refine ⟨Z, hZS.1, fun W hW => ?_⟩
by_cases hWc : matrixSdpDualObjective model W ≤ c
· exact hZmin ⟨hW, hWc⟩
· have hcW : c ≤ matrixSdpDualObjective model W := le_of_not_ge hWc
exact hZS.2.trans hcW
/-- The feasible set of the canonical primal SDP is compact. -/
theorem matrixSdpCanonicalPrimalFeasible_isCompact
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
IsCompact
{X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) |
MatrixSdpCanonicalPrimalFeasible params model X} :=
Metric.isCompact_of_isClosed_isBounded
(matrixSdpCanonicalPrimalFeasible_isClosed params model)
(matrixSdpCanonicalPrimalFeasible_isBounded params model)
/-- The canonical primal objective is continuous. -/
theorem continuous_matrixSdpCanonicalPrimalObjective
(params : Parameters) [FieldModel params.q]
(model : MatrixSdpRealization params) :
Continuous fun X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) =>
Complex.re (Matrix.trace (matrixSdpCanonicalObjectiveOperator params model * X)) := by
have hmul :
Continuous fun X : MatrixOperator (matrixSdpCanonicalBlockHilbertSpace params model) =>
matrixSdpCanonicalObjectiveOperator params model * X :=
continuous_const.mul continuous_id
exact Complex.continuous_re.comp hmul.matrix_trace
end MIPStarRE.LDT.SelfImprovement
end
end LIDT
Source