In extensive-width networks, features are recovered sequentially through sharp phase transitions, yielding an effective width k_c that unifies Bayes-optimal generalization error scaling as Θ(k_c d / n).
Var(B)≲ X i̸=j v2 i v2 j Var(Q3 ij), This improves upon the pessimistic upper bound we used in the dense readout, interpolation regime
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
other 1
citation-polarity summary
fields
stat.ML 1years
2026 1verdicts
UNVERDICTED 1roles
other 1polarities
unclear 1representative citing papers
citing papers explorer
-
Sharp feature-learning transitions and Bayes-optimal neural scaling laws in extensive-width networks
In extensive-width networks, features are recovered sequentially through sharp phase transitions, yielding an effective width k_c that unifies Bayes-optimal generalization error scaling as Θ(k_c d / n).