claimroutine
Randomizing good fibers keeps acceptance 2eta
Randomizing values on a small bad set keeps acceptance ≥
first step of the contradiction argument
Claim·§5.4 Obtaining many useful row fibers·inside lemGood row advice is common and gives short answer sets
Statement (paper)
Fix for which the surrogate acceptance is at least , and write . Let be the union, in the matrix domain, of all good row fibers over every possible , and suppose its uniform measure is less than . If is modified at the entries of by replacing each by an independent uniform element of , then the acceptance changes by at most , so the modified function has acceptance at least .
In plain English
Fix with surrogate acceptance at least and let . If the union of all good row fibers has measure below , then replacing on by independent uniform elements of changes the acceptance by at most , so the modified function has acceptance at least .
Why it's here
This is the first step of a contradiction argument: scrambling the values on the small set barely hurts the test, because the acceptance is an average over matrices and only the mass of can be affected, each entry by at most a factor of 2. The modified function still has acceptance at least (from ), which is what the later shortcode-contradiction claim needs, while good fibers are now fully randomized and so carry little agreement.}}</br>Correction: the JSON above must be a single valid object.</br>{
Classified routine:
A standard union-bound/averaging step: the test queries two (or a bounded number of) points, each uniformly distributed, so modifying f on a set of measure Pr[B] changes the acceptance by at most 2Pr[B].
Uses
Nothing else in the paper
Used by
Sources
PaperOpen
- Section
- §5.4 Obtaining many useful row fibers
- Page
- —
- Label
—- Line
sections/04-matrix.tex:357
Leannot aligned
Phase 2 aligns this node to a Lean declaration.