BatchNormalization.md (403B)
1 # Batch Normalization 2 3 ML P569 4 5 **Definition:** Batch normalization is the process of adding layers to a neural network that perform normalization upon inputs and output the normalized values. 6 7 This helps with unstable gradient issues and removes the need to normalize inputs for the network. On the flip side, these computations are bad for TPUs and are generally slow. They also don't work with RNNs.