Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift

Short summary

Batch Normalization, a new technique that normalizes layer inputs per mini-batch, accelerates deep network training by reducing internal covariate shift, enabling 14x fewer training steps for state-of-the-art image classification models and improving ImageNet accuracy to surpass human raters.

AI-generated from the title and abstract; the full text is not read.