Layer Normalization

A neural network normalization technique that standardizes activations to zero mean and unit variance across the feature dimension for each sample. It stabilizes activation values across layers and is widely used to improve training stability in Transformer and other sequence models.