Coarse-graining support: #
Support.Book.Ch05.Theorems.Section54.VarianceBoundGoodScale.RefinedProbeMoments
Imported from the Apache-2.0 CoarseGraining development at commit
c7ddd76c08ade64fed1b8d2ca51be14dfee8deb4.
Refined probe moments at a good scale #
The coarse probe moment bounds in QuadraticProbeBounds dominate every probe
by one scalar times Λ + λ⁻¹. For the final good-scale variance estimate we
need the matched form: upper coordinates cost only the upper unit-scale factor
and lower coordinates cost only the lower inverse factor. This file supplies
that local refinement and compresses it to \widetilde\Theta_0 using the two
good-scale scalar hypotheses.
Upper unit-scale coefficient carried by a coordinate probe after
normalization at scale m. It is nonzero only on the upper block.
Equations
- HCPolySupport.Book.Ch05.Section54.VarianceBoundGoodScale.coordinateProbeUpperCoeffAtScale hP hStruct m (Sum.inl val) = (hP.barSigmaAtScale hStruct ↑m)⁻¹
- HCPolySupport.Book.Ch05.Section54.VarianceBoundGoodScale.coordinateProbeUpperCoeffAtScale hP hStruct m (Sum.inr val) = 0
Instances For
Lower unit-scale coefficient carried by a coordinate probe after
normalization at scale m. It is nonzero only on the lower block.
Equations
- HCPolySupport.Book.Ch05.Section54.VarianceBoundGoodScale.coordinateProbeLowerCoeffAtScale hP hStruct m (Sum.inl val) = 0
- HCPolySupport.Book.Ch05.Section54.VarianceBoundGoodScale.coordinateProbeLowerCoeffAtScale hP hStruct m (Sum.inr val) = hP.barSigmaStarAtScale hStruct ↑m
Instances For
Matched pointwise domination for coordinate probes: upper coordinates see only the upper unit-scale factor, lower coordinates only the lower inverse factor.
Matched pointwise domination for plus probes.
Matched pointwise domination for minus probes.
Centered origin moment for coordinate probes, compressed to
\widetilde\Theta_0 by the good-scale hypotheses.
Centered origin moment for plus probes, compressed to
\widetilde\Theta_0 by the good-scale hypotheses.
Centered origin moment for minus probes, compressed to
\widetilde\Theta_0 by the good-scale hypotheses.