Adjusting for Dropout Variance in Batch Normalization and Weight Initialization

Abstract

We show how to adjust for the variance introduced by dropout with corrections to weight initialization and Batch Normalization, yielding higher accuracy. Though dropout can preserve the expected input to a neuron between train and test, the variance of the input differs. We thus propose a new weight initialization by correcting for the influence of dropout rates and an arbitrary nonlinearity's influence on variance through simple corrective scalars. Since Batch Normalization trained with dropout estimates the variance of a layer's incoming distribution with some inputs dropped, the variance also differs between train and test. After training a network with Batch Normalization and dropout, we simply update Batch Normalization's variance moving averages with dropout off and obtain state of the art on CIFAR-10 and CIFAR-100 without data augmentation.

References in corpus (6)

• Fast and Accurate Deep Network Learning by Exponential Linear Units (ELUs)
• Densely Connected Convolutional Networks
• Wide Residual Networks
• Weight Normalization: A Simple Reparameterization to Accelerate Training of Deep Neural Networks
• Layer Normalization
• All you need is a good init

Cited by in corpus (6)

• Gaussian Error Linear Units (GELUs)
• Regularization for Deep Learning: A Taxonomy
• Effective and Efficient Dropout for Deep Convolutional Neural Networks
• SUNLayer: Stable denoising with generative networks
• Fire SSD: Wide Fire Modules based Single Shot Detector on Edge Device
• Revisiting Initialization of Neural Networks