This paper presents a statistical framework that yields exact predictions for feature learning and generalization in finite-width, deep, nonlinear networks in the interpolation regime. Learn more: https://go.aps.org/4yC2JnR

0 likes 0 replies