Inherent Noise in Gradient Based Methods

Arushi Gupta

Previous work has examined the ability of larger capacity neural networks to generalize better than smaller ones, even without explicit regularizers, by analyzing gradient based al…