Skip to yearly menu bar Skip to main content


How Does Batch Normalization Help Optimization?

Shibani Santurkar · Dimitris Tsipras · Andrew Ilyas · Aleksander Madry

Room 210 #10

Keywords: [ Deep Learning ] [ Optimization for Deep Networks ] [ Efficient Training Methods ]


Batch Normalization (BatchNorm) is a widely adopted technique that enables faster and more stable training of deep neural networks (DNNs). Despite its pervasiveness, the exact reasons for BatchNorm's effectiveness are still poorly understood. The popular belief is that this effectiveness stems from controlling the change of the layers' input distributions during training to reduce the so-called "internal covariate shift". In this work, we demonstrate that such distributional stability of layer inputs has little to do with the success of BatchNorm. Instead, we uncover a more fundamental impact of BatchNorm on the training process: it makes the optimization landscape significantly smoother. This smoothness induces a more predictive and stable behavior of the gradients, allowing for faster training.

Live content is unavailable. Log in and register to view live content