claimroutineinferred_dep
High-rank families detect with probability 3/16
Families of rank ≥ detect the noise with probability ≥ 3/16
gives the detection property of the lemma
Statement (paper)
Let be any linear subspace with . Then , where , and for each remaining copy the subspace detects with probability at least . Hence , which is Equation , for every .
In plain English
If has , then for a random injection the pullback equals except with probability at most . For every other copy, detects with probability at least , so detects under with probability at least (for ).
Why it's here
This closes the proof of the parent lemma by verifying the detection property in Equation (latent-detection). A high-rank restricts onto all of for most injections , and for those copies the full-lift detection proposition gives probability at least . Combining the fraction of good copies with this gives , with the threshold fixed independently of .
Classified routine:
A direct combination of the full-lift detection proposition with a standard random-injection rank bound: a random linear map pulls back an r-dimensional subspace of K^* onto all of B^* except with probability at most 2^{b-r}, giving 3/4 · 1/4 = 3/16.
Uses
Used by
Nothing in the paper
Sources
PaperOpen
- Section
- §3.5 Quotients and enlargement
- Page
- —
- Label
—- Line
sections/02-latent.tex:711
Leannot aligned
Phase 2 aligns this node to a Lean declaration.
Friction
- infoinferred_depoutline: "High-rank families restrict onto B and detect" rests on prop:latent-full-lift-detection
- infoinferred_depoutline: "High-rank families restrict onto B and detect" rests on claim-lem-latent-2