Section 11 commutativity: scalar marginalization lemmas #
SchwartzβZippel marginalization helpers (eq:evaluate-gcom-at-points, eq:gcom4-diff) used in
the final full-slice commutation theorem: the counterpart of
MIPRE/Background/LIDT/MIPStarRE/LDT/Commutativity/Main/Auxiliary/ScalarMarginalization.lean in
the port of planning/c6b-plan.md (milestone M7, section "Port conventions").
The public lemmas fullSlice_scalar_marginalize_x and fullSlice_scalar_marginalize_y are pure
scalar inequalities. Their proofs compose the tensor-form comparisons over the tensor averages of
Co/Commutativity/Transport/FullSlice/Averages.lean, with closenessOfIP at cost βΞΆ each (the
scalar/tensor decision is recorded upstream in docs/decisions/713-scalar-tensor-decision.md).
The vendored hypothesis hnorm : strategy.state.IsNormalized is dropped from both public
lemmas: normalization is a theorem of the model (VecState.ev_one_of_isNormalized), and the
ported Preliminaries.switchSandwich and transport bridges no longer take it. The two
summation helpers take the model S : SymModel π K as an explicit first argument, where the
vendored ones take Ο; avgOver_slice_total_left_sandwich_eq takes it after params, as the
vendored one takes Ο.
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/commutativity-G.texblueprint/src/chapter/ch08_commutativity.tex
The evaluated point family viewed as a projective submeasurement family.
Equations
- MIPRE.LIDT.Co.Commutativity.evaluatedPointProj params family u = { toSubMeas := MIPRE.LIDT.Co.Commutativity.evaluatedPointFamily params family u, proj := β― }
Instances For
Summing the inner outcome in an ABA expectation turns it into the
submeasurement total.
Summing the right-register outcome in the middle switch-sandwich term turns it into the submeasurement total.
Averaging the middle total in an ABA sandwich produces the averaged
slice operator G = E_y GΚΈ.
Full-slice cubic first term as the left switch-sandwich expectation.
Evaluated-slice cubic first term as the evaluated left switch-sandwich expectation.
The full and evaluated switch-sandwich middle terms are the same G β G average.
Paper first-term switch-sandwich transport
(commutativity-G.tex lines 295--305), stated in the public scalar API.
The paper does not use an md/q Schwartz--Zippel step for the cubic first term.
Instead, both the full and evaluated cubic terms are compared to the common
G β G switch-sandwich center, costing 2βΞΆ on each side.
Paper-faithful second-term transport bound.
The proved x-prefix (eq:gcom4 plus eq:gcom4-diff, paper lines 332--354)
costs md/q + βΞΆ; the proved line-359 closenessOfIP comparison costs βΞΆ;
the line-360 scalarβtensor comparison is proved in
xEvaluatedFullSliceABABAvg_to_xEvaluatedFullSliceABABtensorAvg and costs
another βΞΆ; and the proved y-tail uses y-Schwartz--Zippel marginalization
(paper lines 369--385) plus the βΞΆ doubly-evaluated scalarβtensor comparison. Thus
the whole scalar second-term comparison costs 2Β·md/q + 4βΞΆ.