Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
At a glance
Batch normalization rescales signals inside neural networks to speed up training.
In one paragraph
Introduced batch normalization, adjusting the scale of signals inside neural networks during training, allowing faster learning with larger updates and less sensitivity to initial settings.
