Documentation

MIPRE.Background.LIDT.MIPStarRE.LDT.Pasting.Bernoulli.Final

Section 12 pasting: final pasting theorems #

Final completeness and pasting theorems.

theorem MIPStarRE.LDT.Pasting.ldPastingNCompleteness_of_overAllOutcomes_fromHToG_tail {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (heps_nonneg : 0 eps) (hdelta_nonneg : 0 delta) (hgamma_nonneg : 0 gamma) (hzeta_nonneg : 0 zeta) (family : IdxPolyFamily params ι) (k : ) (hk_pos : 1 k) (hOAO : OverAllOutcomesStatement params strategy family eps delta gamma zeta k) (hFrom : FromHToGStatement params strategy strategy.state family gamma zeta k) (htail : 1 - kappa * (1 + 1 / (100 * params.m)) - Real.exp (-(k / (80000 * params.m ^ 2))) fromHToGBernoulliTailMass params strategy.state family k) :
LdPastingNCompletenessStatement params strategy family kappa (MainInductionStep.ldPastingInInductionNu params k eps delta gamma zeta) k

Internal form of cor:ld-pasting-N-completeness from the two preceding mass-comparison inputs.

This theorem isolates the scalar assembly after lem:over-all-outcomes, lem:from-H-to-G, and the Bernoulli-tail lower bound have already been established.

theorem MIPStarRE.LDT.Pasting.ldPastingNCompleteness_of_tailLowerBound {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (hgamma_le : gamma 1) (hzeta_le : zeta 1) (hdq_le : params.d params.q) (hd : 0 < params.d) (family : IdxPolyFamily params ι) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (hk_pos : 1 k) (_hk : 400 * params.m * params.d k) (htail : 1 - kappa * (1 + 1 / (100 * params.m)) - Real.exp (-(k / (80000 * params.m ^ 2))) fromHToGBernoulliTailMass params strategy.state family k) :
LdPastingNCompletenessStatement params strategy family kappa (MainInductionStep.ldPastingInInductionNu params k eps delta gamma zeta) k

cor:ld-pasting-N-completeness once the Bernoulli-tail lower bound is supplied explicitly.

This records the downstream scalar algebra after lem:over-all-outcomes and lem:from-H-to-G. The hypothesis htail is exactly the θ = 1 / (200m) specialization of lem:chernoff-bernoulli-matrix for the averaged complete operator G = \mathbb E_x \sum_g G^x_g, expressed as the concrete fromHToGBernoulliTailMass lower bound with error κ · (1 + 1/(100m)) + exp(-k / (80000 m²)).

theorem MIPStarRE.LDT.Pasting.ldPastingNCompleteness {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (hgamma_le : gamma 1) (hzeta_le : zeta 1) (hdq_le : params.d params.q) (hd : 0 < params.d) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (hk_pos : 1 k) (hk : 400 * params.m * params.d k) :
LdPastingNCompletenessStatement params strategy family kappa (MainInductionStep.ldPastingInInductionNu params k eps delta gamma zeta) k

cor:ld-pasting-N-completeness.

theorem MIPStarRE.LDT.Pasting.ldPastingNCompleteness_ofComMain_of_axis_self {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (haxis : strategy.axisParallelFailureProbability eps) (hself_good : strategy.selfConsistencyFailureProbability delta) (hgamma_nonneg : 0 gamma) (hgamma_le : gamma 1) (hzeta_nonneg : 0 zeta) (hzeta_le : zeta 1) (hdq_le : params.d params.q) (hd : 0 < params.d) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hcom : Commutativity.ComMainConclusion params strategy family gamma zeta) (k : ) (hk_pos : 1 k) (hk : 400 * params.m * params.d k) :
LdPastingNCompletenessStatement params strategy family kappa (MainInductionStep.ldPastingInInductionNu params k eps delta gamma zeta) k

Internal form of cor:ld-pasting-N-completeness from the Section 11 commutativity conclusion.

theorem MIPStarRE.LDT.Pasting.ldPastingSubMeas {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (hgamma_le : gamma 1) (hzeta_le : zeta 1) (hdq_le : params.d params.q) (hd : 0 < params.d) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (hk_pos : 1 k) (hk : 400 * params.m * params.d k) :
∃ (H : SubMeas (Polynomial params.next) ι), H = constructedPastedSubMeas params family k LdPastingSubMeasConclusion params strategy family H eps delta gamma kappa zeta k

lem:ld-pasting-sub-measurement.

theorem MIPStarRE.LDT.Pasting.ldPastingNontrivial {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (hgamma_le : gamma 1) (hzeta_le : zeta 1) (hdq_le : params.d params.q) (hd : 0 < params.d) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (hk_pos : 1 k) (hk : 400 * params.m * params.d k) :
∃ (H : Measurement (Polynomial params.next) ι), H = constructedPastedMeasurement params family k LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Restricted nontrivial-regime Lean form of thm:ld-pasting.

The source theorem is references/ldt-paper/ld-pasting.tex, lines 12--50. Lines 52--55 explain that the proof may assume the nontrivial regime eps, delta, gamma, zeta, d / q ≤ 1, since the complementary cases are trivial. This declaration states the restricted assumptions gamma ≤ 1, zeta ≤ 1, params.d ≤ params.q, 0 < params.d, and 1 ≤ k. The unrestricted statement aligned with the paper is ldPasting; the degree-zero complementary branch is handled separately by ldPastingDegreeZeroBranch.

theorem MIPStarRE.LDT.Pasting.ldPasting_of_one_le_error {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (family : IdxPolyFamily params ι) (k : ) (herror : 1 MainInductionStep.ldPastingInInductionError params k eps delta gamma kappa zeta) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Trivial consistency conclusion when the target pasting error is at least 1.

The consistency defect of two submeasurements against a normalized bipartite state is always at most 1; hence a scalar lower bound 1 ≤ ldPastingInInductionError ... is enough to produce the final conclusion with a distinguished trivial measurement.

theorem MIPStarRE.LDT.Pasting.ldPasting_of_one_le_nu_or_zero_k {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (k : ) (hnu : 1 k1 MainInductionStep.ldPastingInInductionNu params k eps delta gamma zeta) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Trivial consistency conclusion from the complementary scalar branches.

If k is positive, it suffices to show that the ν term in the pasting error is at least 1; if k = 0, the exponential term already gives the trivial bound.

theorem MIPStarRE.LDT.Pasting.ldPastingLargeGammaBranch {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (_hself : family.StronglySelfConsistent strategy.state zeta) (_hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (_hk : 400 * params.m * params.d k) (hgamma : 1 < gamma) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Complementary branch for thm:ld-pasting when gamma > 1.

Paper origin: references/ldt-paper/ld-pasting.tex:52-55, where this is one of the large-error cases in which the final consistency bound is trivial. This is one of the proved complementary cases for thm:ld-pasting.

theorem MIPStarRE.LDT.Pasting.ldPastingLargeZetaBranch {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (_hcons : family.ConsistentWithPoints strategy zeta) (_hself : family.StronglySelfConsistent strategy.state zeta) (_hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (_hk : 400 * params.m * params.d k) (hzeta : 1 < zeta) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Complementary branch for thm:ld-pasting when zeta > 1.

Paper origin: references/ldt-paper/ld-pasting.tex:52-55, where this is one of the large-error cases in which the final consistency bound is trivial. This is one of the proved complementary cases for thm:ld-pasting.

theorem MIPStarRE.LDT.Pasting.ldPastingLargeDegreeRatioBranch {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (_hself : family.StronglySelfConsistent strategy.state zeta) (_hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (_hk : 400 * params.m * params.d k) (hdq : params.q < params.d) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Complementary branch for thm:ld-pasting when d > q.

Paper origin: references/ldt-paper/ld-pasting.tex:52-55, where this is the large-error case (d/q) ≥ 1. This is one of the proved complementary cases for thm:ld-pasting.

theorem MIPStarRE.LDT.Pasting.ldPastingZeroKBranch {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (_hgood : strategy.IsGood eps delta gamma) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (_hcons : family.ConsistentWithPoints strategy zeta) (_hself : family.StronglySelfConsistent strategy.state zeta) (_hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (_hk : 400 * params.m * params.d k) (hk_zero : k = 0) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Complementary branch for thm:ld-pasting when k = 0.

This branch is a boundary case for the reduction to the nontrivial theorem, whose proof assumes 1 ≤ k. The scalar calculation showing that the exponential term gives the trivial bound is proved in ScalarBounds.lean.

theorem MIPStarRE.LDT.Pasting.ldPastingDegreeZeroBranch {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (_hself : family.StronglySelfConsistent strategy.state zeta) (_hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (_hk : 400 * params.m * params.d k) (hd_zero : params.d = 0) (_hk_pos : 1 k) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Degree-zero complementary branch for the unrestricted source theorem.

Paper origin: references/ldt-paper/ld-pasting.tex:12-55. The paper's large-error reduction names the cases eps, delta, gamma, zeta, d/q ≥ 1; it does not explicitly add 0 < d as a hypothesis of thm:ld-pasting. Thus the Lean theorem should not add 0 < d as an assumption of that cited theorem.

Issue #1622 recorded the need for a direct proof of this degree-zero branch; see docs/paper-gaps/issue-1622-ld-pasting-degree-zero.tex. The existing nontrivial argument cannot simply be reused: its hBConsistency aggregation passes from distinct sampled heights to independent sampled heights and absorbs the resulting k^2/q loss through the displayed (d/q)^(1/32) term. When d = 0, that term is zero, so the branch requires a separate argument rather than an additional hypothesis on ldPasting.

theorem MIPStarRE.LDT.Pasting.ldPastingNontrivialPublicBranch {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (hgamma_le : gamma 1) (hzeta_le : zeta 1) (hdq_le : params.d params.q) (hd : 0 < params.d) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (hk_pos : 1 k) (hk : 400 * params.m * params.d k) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Projection from the restricted nontrivial construction.

The restricted construction theorem ldPastingNontrivial proves the nontrivial analytic regime for the canonical pasted measurement. This auxiliary statement records the projection from the restricted construction theorem to the conclusion needed by the unrestricted theorem, without changing the statement of thm:ld-pasting.

theorem MIPStarRE.LDT.Pasting.ldPasting {ι : Type u_1} [Fintype ι] [DecidableEq ι] (params : Parameters) [FieldModel params.q] (strategy : SymStrat params.next ι) (eps delta gamma kappa zeta : Error) (hgood : strategy.IsGood eps delta gamma) (family : IdxPolyFamily params ι) (hcomplete : family.Complete strategy.state kappa) (hcons : family.ConsistentWithPoints strategy zeta) (hself : family.StronglySelfConsistent strategy.state zeta) (hbound : IdxPolyFamily.SliceBoundednessInput strategy family zeta) (k : ) (hk : 400 * params.m * params.d k) :
∃ (H : Measurement (Polynomial params.next) ι), LdPastingConclusion params strategy family H eps delta gamma kappa zeta k

Paper-aligned form of thm:ld-pasting.

Paper origin: references/ldt-paper/ld-pasting.tex, lines 12--50. The following lines 52--55 explain that the proof may restrict to the regime eps, delta, gamma, zeta, d / q ≤ 1, because the complementary cases are trivial. The restricted theorem ldPastingNontrivial proves the nontrivial regime, and the large-gamma, large-zeta, large-d / q, and k = 0 complementary branches are proved above, including the degree-zero case, so this declaration keeps the unrestricted paper statement visible without adding the non-paper assumptions from the restricted theorem. The former obstruction is documented in docs/paper-gaps/issue-1622-ld-pasting-degree-zero.tex.