Documentation

MIPStarRE.LDT.SelfImprovement.Theorems.Results.BoundednessTransport.Decomposition

Boundedness transport decomposition identities #

This file contains the algebraic reindexing and off-diagonal decompositions used in the final-fields boundedness and point-consistency arguments. The identities come from the helper-stage agreement average in the proof of self-improvement.

References #

theorem MIPStarRE.LDT.SelfImprovement.helper_agreement_average_ev_eq_avg {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params ι) (H : SubMeas (Polynomial params) ι) :
ev strategy.state (helperAgreementAverageOperator params strategy H) = avgOver (uniformDistribution (Point params)) fun (u : Point params) => a : Fq params, ev strategy.state (opTensor ((strategy.pointMeasurement u).outcome a) ((evaluateAt params u H).outcome a))

Scalar expansion of the averaged helper-agreement operator.

The expectation of the averaged agreement operator is the average, over the point question, of the scalar agreement between the point measurement and the postprocessed polynomial family.

theorem MIPStarRE.LDT.SelfImprovement.helperAgreementOperatorAtPoint_eq_sum_polynomial {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params ι) (H : SubMeas (Polynomial params) ι) (u : Point params) :
helperAgreementOperatorAtPoint params strategy H u = h : Polynomial params, opTensor ((strategy.pointMeasurement u).outcome (h.toFun u)) (H.outcome h)

Reindexing identity for the pointwise helper-agreement operator.

The fiberwise definition H_{[h(u)=a]} := ∑_{h : h(u)=a} H_h collapses the a-summed expression ∑_a A^u_a ⊗ H_{[h(u)=a]} to the polynomial-indexed sum ∑_h A^u_{h(u)} ⊗ H_h, by expanding the tensor product fiberwise and applying Finset.sum_fiberwise along h ↦ h u.

This is the first equality of the boundedness display in the proof of \ref{item:self-improvement-boundedness}: references/ldt-paper/self_improvement.tex line 612, mirrored at blueprint/src/chapter/ch07_self_improvement.tex lines 274--282 ("Reindexing the sum by~$h$"). It is a purely algebraic identity — no estimate, no measurement structure used beyond the postprocess fiber decomposition built into evaluateAt.

theorem MIPStarRE.LDT.SelfImprovement.helper_agreement_average_ev_eq_polynomial_sum {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params ι) (H : SubMeas (Polynomial params) ι) :
ev strategy.state (helperAgreementAverageOperator params strategy H) = avgOver (uniformDistribution (Point params)) fun (u : Point params) => h : Polynomial params, ev strategy.state (opTensor ((strategy.pointMeasurement u).outcome (h.toFun u)) (H.outcome h))

Reindexed expansion of the averaged helper-agreement operator.

Combining the pointwise reindexing identity helperAgreementOperatorAtPoint_eq_sum_polynomial with helper_agreement_average_ev_eq_avg, the scalar ⟨ψ| E_u Σ_a A^u_a ⊗ H_{[h(u)=a]} |ψ⟩ equals the polynomial-indexed expectation E_u Σ_h ⟨ψ| A^u_{h(u)} ⊗ H_h |ψ⟩ from the second line of the boundedness display in the proof of \ref{item:self-improvement-boundedness} (references/ldt-paper/self_improvement.tex line 612; blueprint/src/chapter/ch07_self_improvement.tex lines 274--282).

theorem MIPStarRE.LDT.SelfImprovement.helperAgreementOperatorAtPoint_off_diagonal_decomposition {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params ι) (H : SubMeas (Polynomial params) ι) (u : Point params) :
rightTensor H.total - helperAgreementOperatorAtPoint params strategy H u = h : Polynomial params, aFinset.univ.erase (h.toFun u), opTensor ((strategy.pointMeasurement u).outcome a) (H.outcome h)

Off-diagonal decomposition of the pointwise helper boundedness slack.

For each point u, the difference between the right-placed total I ⊗ H.total = ∑_h I ⊗ H_h and the pointwise helper-agreement operator helperAgreementOperatorAtPoint params strategy H u = ∑_a A^u_a ⊗ H_{[h(u)=a]} equals the off-diagonal sum ∑_h ∑_{a ≠ h(u)} A^u_a ⊗ H_h, by combining the polynomial-indexed reindexing of helperAgreementOperatorAtPoint from #1124 (helperAgreementOperatorAtPoint_eq_sum_polynomial) with ∑_a A^u_a = 1 (since pointMeasurement u is a measurement) and the bilinearity of opTensor.

This is the operator-level form of the second algebraic identity in the boundedness display in \ref{item:self-improvement-boundedness} (references/ldt-paper/self_improvement.tex line 613, mirrored at blueprint/src/chapter/ch07_self_improvement.tex lines 296--300, the step "Combined with $\sum_a A_a^u = I$ and~\eqref{eq:explicit-bound-for-A-consistency} this gives ..."). The averaged scalar form of the off-diagonal sum on the right is the LHS of eq:explicit-bound-for-A-consistency (line 435), which the paper bounds by 4 √ζ_variance.

theorem MIPStarRE.LDT.SelfImprovement.helperAgreementOperatorAtPoint_ev_slack_eq_off_diagonal_sum {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params ι) (H : SubMeas (Polynomial params) ι) (u : Point params) :
ev strategy.state (rightTensor H.total) - ev strategy.state (helperAgreementOperatorAtPoint params strategy H u) = h : Polynomial params, aFinset.univ.erase (h.toFun u), ev strategy.state (opTensor ((strategy.pointMeasurement u).outcome a) (H.outcome h))

Scalar form of the pointwise off-diagonal decomposition.

For each evaluation point u, the scalar slack ⟨ψ, I ⊗ H.total, ψ⟩ - ⟨ψ, helperAgreementOperatorAtPoint u, ψ⟩ is the sum of the off-diagonal masses ⟨ψ, A^u_a ⊗ H_h, ψ⟩ over the pairs with a ≠ h(u).

theorem MIPStarRE.LDT.SelfImprovement.helper_boundedness_slack_average_ev_eq_off_diagonal_avg {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params ι) (H : SubMeas (Polynomial params) ι) :
ev strategy.state (rightTensor H.total) - ev strategy.state (helperAgreementAverageOperator params strategy H) = avgOver (uniformDistribution (Point params)) fun (u : Point params) => h : Polynomial params, aFinset.univ.erase (h.toFun u), ev strategy.state (opTensor ((strategy.pointMeasurement u).outcome a) (H.outcome h))

Averaged scalar form of the off-diagonal decomposition.

Composed from helperAgreementOperatorAtPoint_off_diagonal_decomposition by applying the bilinearity of ev/avgOver over subtraction and averaging via avgOver_uniform_const. The difference ⟨ψ, I ⊗ H.total, ψ⟩ - ⟨ψ, helperAgreementAverageOperator, ψ⟩ equals the averaged off-diagonal scalar sum E_u ∑_h ∑_{a ≠ h(u)} ⟨ψ, A^u_a ⊗ H_h, ψ⟩, which is the LHS of eq:explicit-bound-for-A-consistency (references/ldt-paper/self_improvement.tex line 435; blueprint ch07_self_improvement.tex lines 153--168).