Fast generalization error bound of deep learning without scale invariance of activation functions

Yoshikazu Terada, Ryoma Hirose

In theoretical analysis of deep learning, discovering which features of deep learning lead to good performance is an important task. In this paper, using the framework for analyzin…