L1 sandwiching suffices for efficient PQ learning via Iterative Chow Filtering, giving quasipolynomial-time algorithms for DNFs under uniform distribution and exponential improvements for circuits and PTFs.
First, since|Strain|is sufficiently large, by Theorem E.1, we can ensure that with probability at least 1−δ/2, it holdsPr(x,y)∼Dtrain[ ˆf(x)̸=y]≤opttrain +ϵη/2
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DS 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Iterative Chow Filtering for Learning with Distribution Shift
L1 sandwiching suffices for efficient PQ learning via Iterative Chow Filtering, giving quasipolynomial-time algorithms for DNFs under uniform distribution and exponential improvements for circuits and PTFs.