1 paper · 1 filter
Paolo Baglioni, Christian Keup, Vincenzo Zimbardo +4
The scaling limit where both the size of the training set P and the width N of a deep neural network grow at the same rate, the so-called proportional-width regime, has been in…