Paper
Mesoscopic learning dynamics of extensive-width neural networks
A paper that studies the learning dynamics of wide neural networks, proposes a mesoscopic description, and reveals the mechanism of feature specialization and the phenomenon of acceleration by quadratic screening.
Timeline (last 90 days)
Oct 19
The paper 'Mesoscopic learning dynamics of extensive-width neural networks' studies a two-layer neural network learning a target function represented by a network of the same architecture, where the network width and input dimension grow proportionally.
Unverified50%