$\color{black}\rule{365px}{3px}$

“Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift” - 2015

Link to Paper:

arxiv.org

“How Does Batch Normalization Help Optimization?” - 2019

Link to Paper:

arxiv.org

Table of Contents

REMINDER

Batch Norm works well and provides substantial measurable benefits to deep learning architecture design and training. However, there is still no universally agreed answer about what gives it its amazing powers.

1. Batch Normalization

$\color{black}\rule{365px}{3px}$