Quasipolynomial-time algorithms learn AC^0 circuits under graphical models with polynomial growth and strong spatial mixing by transferring low-degree approximations via new sampling methods.
Smoothed Agnostic Learning of Halfspaces over the Hypercube
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Learning $\mathsf{AC}^0$ Under Graphical Models
Quasipolynomial-time algorithms learn AC^0 circuits under graphical models with polynomial growth and strong spatial mixing by transferring low-degree approximations via new sampling methods.