claimroutine
Coarse bin scale factor b/a
For the first-channel scale factor is
coarse-bin case of the scale factor
Claim·§7.3 Mean comparisons and the cost of releasing a source·inside lemMean comparison scale factors and release cost
Statement (paper)
For , the bound for the th-power norm of a first-channel mean contribution (its clipped relative error against ) carries, on each child root-discrepancy power, the scale factor . This follows from Equation after conditional expectation and the mixture contraction of Lemma , with homogeneity giving a factor after division by . The factor in a product is bounded above and below by fixed constants, and for bounded nonnegative atoms a product-root difference splits into the factor-root differences multiplied by fixed constants times support indicators of the unaffected factor in one of the two versions.
In plain English
When , the bound on the th-power norm of a first-channel mean contribution (its clipped relative error against ) carries scale factor on each child root-discrepancy power.
Why it's here
This is the case of the scale factor in the parent lemma. It follows from the relative-loss equation after conditional expectation and the mixture contraction in the estimation-loss lemma, with homogeneity supplying the after division by . Bounded factors and support indicators split product-root differences into factor-root differences, so no independence of the estimated factors is needed.
Classified routine:
A case of the parent lemma's scale-factor bound. It follows by conditional expectation, the mixture contraction of the estimation loss (lem-est-loss) and homogeneity, so it is bookkeeping on earlier nodes.
Sources
PaperOpen
- Section
- §7.3 Mean comparisons and the cost of releasing a source
- Page
- —
- Label
—- Line
sections/estimation.tex:249
Leannot aligned
Phase 2 aligns this node to a Lean declaration.