How initial conditions affect generalization performance in large networks
{{output}}
Generalization is one of the most important problems in neural-network research. It is influenced by several factors in the network design, such as network size, weight decay factor, and others. We show here that the initial weight distribution (for gradient d... ...
