arXiv 1607.06450
Layer Normalization
By Jimmy Lei Ba, Jamie Ryan Kiros, et al.
Published 2016-07-21
Wiki summary
Explore the paper's summary, context, and related research on Papiers.
Training state-of-the-art, deep neural networks is computationally expensive. One way to reduce the training time is to normalize the activities of the neurons. A recently introduced technique called batch normalization uses the distribution of the summed input to a neuron over a mini-batch of training cases to compute a mean and variance which are then used to normalize the summed input to that neuron on each train…