Full-slice tensor marginalization core #
Collision residuals, postprocessing expansions, and tensor marginalization
core bounds for the BABA and ABAB full-slice tensor averages.
Ex-private definitions are tensor-form machinery per architecture decision #713; downstream code should use the scalar public API exposed by the full-slice transport theorems.
References #
references/ldt-paper/commutativity-G.texblueprint/src/chapter/ch08_commutativity.tex
Factored collision residual for the y-marginalization tensor step.
Here the outer sandwich is the already x-evaluated family
G^x_[g(u)=a], while the colliding polynomial pair is on the y side.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Averaged x-collision bound in the form consumed by
fullSliceBABA_tensor_marginalize_x.
Averaged y-collision bound in the form consumed by
fullSliceABAB_tensor_marginalize_y. This is the y-side analogue of
fullSliceBABA_tensor_marginalize_x_collision_bound.
Expand one postprocessed tensor sandwich and split the resulting pair sum into its diagonal part and off-diagonal collision residual.
This is the common finite-sum identity behind both tensor marginalization steps.
The outcome family A is postprocessed by the sample-dependent map eval s; the
outer sandwich family B is not postprocessed.
The y-collision residual is nonnegative term-by-term.
X-side tensor marginalization bound for paper eq:gcom4-diff.
This staged statement compares the full BAB ⊗ A tensor average to the
intermediate where only the x polynomial outcome has been evaluated at u.
It is the Lean-local tensor form of the Schwartz-Zippel step labelled
eq:gcom4-diff in the proof of blueprint theorem thm:com-main.