Near n ~ d^2, Bayes-optimal two-layer networks with generic activations and weights have a universal phase, a specialisation phase, and a first-order transition between them.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Optimal generalisation and learning transition in extensive-width shallow neural networks near interpolation
Near n ~ d^2, Bayes-optimal two-layer networks with generic activations and weights have a universal phase, a specialisation phase, and a first-order transition between them.