Batch Normalization
In plain terms
Normalizing layer inputs to stabilize and speed up training.
Going deeper
Batch normalization standardizes the inputs to each layer across a mini-batch, smoothing and accelerating neural-network training. It reduces internal covariate shift.
Under the hood
Allows higher learning rates and adds slight regularization; behaves differently in train vs. inference.
Examples
Deep network training, faster convergence, CNNs, stabilizing gradients.
Think of it like...
Like recalibrating instruments before each step so measurements stay consistent.