A technique that normalizes inputs to neural network layers by adjusting and scaling activations. Batch normalization accelerates training, reduces sensitivity to initialization, acts as regularization, and enables higher learning rates, improving overall model training stability and performance.
