Section 12 pasting: vertical-line consistency transfer #
The ldGbcon transfer compares the slice family G^x with the vertical-line answers B^u. It
combines the conditioned axis-parallel consistency estimate with the point-to-vertical-line
state-dependent-distance bound: the counterpart of
MIPRE/Background/LIDT/MIPStarRE/LDT/Pasting/Core/LdGbcon.lean in the port of
planning/c6b-plan.md (milestone M11, section "Port conventions").
A strategy is a SymStrat params.next ๐ K and the slice family an IdxPolyFamily params ๐, so
the two vertical-line measurements are local families in ๐, and the distance bounds are
SDDRels of their right placements on strategy.state. The vendored swap
consRel_symm_of_density_fixed strategy.state strategy.densityFixed is the model's theorem
SymModel.consRel_symm_of_density_fixed, and the normalization strategy.isNormalized that the
vendored triangleSub_right takes is not a hypothesis of the ported one, so no swap or
normalization hypothesis appears.
Two proofs are shorter than the vendored ones. pointVerticalLineSdd_of_axis_self drops the
vendored point self-consistency ConsRel, which it computed and did not use. ldGbcon_of_axis_self
is Preliminaries.triangleSub_right applied to the point consistency of the family and to
pointVerticalLineSdd_of_axis_self directly, in place of a second copy of that bound's proof
followed by a rewrite along ldGbconAxisLineMeasurement_eq_verticalLineMeasurement.
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
The axis-parallel line measurement in the last direction through u, read at its base
point: the line answer B^u of the axis-parallel test conditioned on the last direction.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The vertical line through u, based at height zero, read at the height of u: the
vertical-line answer B^u as a measurement.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The two readings of B^u agree, by the rebasing covariance of the axis-parallel
measurement: rebasing the vertical line at the height of u gives the last-direction line
through u.
The point measurements are m'ยทฮต-consistent with the last-direction line answers, where
m' = m + 1 is the number of directions and ฮต bounds the axis-parallel test averaged over all
directions.
Axis/self-consistency form of lem:point-vertical-line-sdd.
The proof of the point-to-vertical-line transfer uses only the axis-parallel test and point self-consistency. The diagonal-line test is not part of this calculation.
Named submeasurement-family form of pointVerticalLineSdd_of_axis_self.
The proof above constructs the vertical-line measurement as a complete
measurement-valued family. The degree-zero estimates use only its underlying
submeasurement family, namely liftedVerticalLineAnswerFamily.
lem:point-vertical-line-sdd.
A good strategy induces a state-dependent distance bound between the point
measurements and the vertical-line (axis-parallel rebased) measurements.
lem:ld-gbcon.
This is the direct consistency transfer from the slice family G^x to the
vertical line answers B^u, obtained by composing the hypothesis
item:ld-pasting-consistency with the conditioned axis-parallel test relation.
lem:ld-gbcon.
This is the direct consistency transfer from the slice family G^x to the
vertical line answers B^u, obtained by composing the hypothesis
item:ld-pasting-consistency with the conditioned axis-parallel test relation.
Named-family form of lem:ld-gbcon.
This is the same consistency transfer as ldGbcon, restated with the two
families used in the degree-zero branch of thm:ld-pasting: the evaluated slice
family family.evaluatedAtNextPoint and the lifted vertical-line family
liftedVerticalLineAnswerFamily. In the degree-zero branch, these are the two
families whose pointwise invariance properties must be combined to control
height dependence.
Named-family form of lem:ld-gbcon.