Rethinking the Usage of Batch Normalization and Dropout in the Training of Deep Neural Networks

TL;DR

This paper introduces an innovative IC layer combining Batch Normalization and Dropout to improve neural network training efficiency by promoting input independence, leading to faster convergence and more stable training.

Contribution

The paper proposes a novel IC layer that integrates Batch Normalization and Dropout to enhance training speed and stability, challenging standard practices in neural network design.

Findings

IC layer outperforms baseline methods on CIFAR and ImageNet datasets.

Training becomes more stable with faster convergence.

Revises common practices by combining Batch Normalization and Dropout.

Abstract

In this work, we propose a novel technique to boost training efficiency of a neural network. Our work is based on an excellent idea that whitening the inputs of neural networks can achieve a fast convergence speed. Given the well-known fact that independent components must be whitened, we introduce a novel Independent-Component (IC) layer before each weight layer, whose inputs would be made more independent. However, determining independent components is a computationally intensive task. To overcome this challenge, we propose to implement an IC layer by combining two popular techniques, Batch Normalization and Dropout, in a new manner that we can rigorously prove that Dropout can quadratically reduce the mutual information and linearly reduce the correlation between any pair of neurons with respect to the dropout layer parameter $p$. As demonstrated experimentally, the IC layer consistently outperforms the baseline approaches with more stable training process, faster convergence speed and better convergence limit on CIFAR10/100 and ILSVRC2012 datasets. The implementation of our IC layer makes us rethink the common practices in the design of neural networks. For example, we should not place Batch Normalization before ReLU since the non-negative responses of ReLU will make the weight layer updated in a suboptimal way, and we can achieve better performance by combining Batch Normalization and Dropout together as an IC layer.