Section 12 pasting: from-H-to-G paper telescope #
Assembles the adjacent-stage paper move chain A → M₁ → M₂ → M₃ → E and records the final
stage-mass telescope for the fromHToG reduction: the counterpart of
MIPRE/Background/LIDT/MIPStarRE/LDT/Pasting/Bernoulli/FromHToG/PaperMoveChain/Telescope.lean in
the port of planning/c6b-plan.md (milestone M11, section "Port conventions").
The vendored bipartite state ψbi : QuantumState (ι × ι) is the symmetric model
S : SymModel 𝔓 K, in the vendored argument position; leftTensor/rightTensor are
S.L/S.R, ᴴ is star, and the vendored weight operator, also named S, is W. The three
lemmas take the normalization hypothesis hnorm : ψbi.IsNormalized, used only to feed
closenessOfIP, closenessOfIPAdjoint and the two paper moves of Moves, whose ported forms
drop it; it is dropped here as well (section "Port conventions"), so vendored callers pass
strategy.state where they passed ψbi hnorm. No lemma has a swap or density hypothesis.
The two Cauchy--Schwarz context bounds (A → M₁ and M₁ → M₂) are proved with the keystone's
placement positivity (S.opTensor_nonneg, S.opTensor_le_leftTensor, S.opTensor_mono_right)
and the projectivity gHatIdxMeas_proj of the completed outcomes, in place of the vendored
Kronecker and sq_le_self chains. They are the two private theorems
fromHToGAdjacentStage_moveA0M1 and fromHToGAdjacentStage_moveM1M2, which the vendored proof of
fromHToGAdjacentStage_paperMoveChain has inline: inline, that proof took 3.6 s, and split, each
takes 1.6–2.3 s (load about 1; the vendored proof, 11.6 s).
Not ported #
Every declaration of the vendored file has a counterpart here.
References #
In LionSR/MIPStarRE at commit 507e8122, not in this repository:
references/ldt-paper/ld-pasting.tex
One adjacent fromHToG paper step.
Adjacent-stage recurrence obtained by applying the paper move chain at every nonterminal stage.
The paper-total stage-mass telescope for fromHToG.
This follows the iteration in ld-pasting.tex:1354--1372: applying the adjacent-stage
estimate over all k stages gives k copies of the per-stage error. Lean records
that literal telescope before the final scalar bound fromHToGPaperTotalError_le
absorbs it into fromHToGError.