Deep Gradient Boosting -- Layer-wise Input Normalization of Neural Networks

Erhan Bilal

Stochastic gradient descent (SGD) has been the dominant optimization method for training deep neural networks due to its many desirable properties. One of the more remarkable and l…