lemmaroutine
Dense-bin atom table is bounded and unbiased
Dense-bin atom table is bounded with conditional mean
unbiased atom estimator behind the correction atom
Lemma·
lem:atom-dense·§5.2 Dense bins and preparatory normalizationStatement (paper)
For sufficiently large fixed , the table
on retained starts has bounded values and row support. On the slice
its conditional expectation times equals
. The same assertion holds for
with the routing fraction included.
Apart from its start membership, its true value uses only the terminal
channel.
In plain English
For large enough fixed , the table on retained starts has bounded values and bounded row support. On the slice , its conditional expectation times equals . The same holds for with the routing fraction included, and apart from start membership its true value depends only on the terminal channel.
Why it's here
This is the basic atom identity for dense bins: it shows the sampled table is a bounded, unbiased estimator of , so later sample-mean arguments can use it. The retention estimates from the rank-channel lemma control the conditioning on the slice. The terminal-channel-only dependence of the true value lets the correction atom lemma and other later steps reuse it.
Classified routine:
A direct unbiasedness and boundedness check for an importance-weighted indicator estimator. It follows by conditioning on the slice and applying the rank-channel retention estimates (lem-env-ranks) to the normalization .
Sources
PaperOpen
- Section
- §5.2 Dense bins and preparatory normalization
- Page
- —
- Label
lem:atom-dense- Line
sections/atoms.tex:224
Leannot aligned
Phase 2 aligns this node to a Lean declaration.