Batch Normalization is a technique to improve learning in neural networks by normalizing the distribution of each input feature in each layer across each minibatch to N(0, 1).
