Just Pick a Sign: Optimizing Deep Multitask Models with Gradient Sign Dropout

TL;DR

This paper introduces Gradient Sign Dropout (GradDrop), a novel probabilistic masking technique for gradient signals in deep multitask models, improving training efficiency and performance by managing conflicting gradient updates.

Contribution

GradDrop is a simple, versatile layer that enhances multitask training by selectively masking gradients based on their consistency, outperforming existing multiloss methods.

Findings

GradDrop outperforms state-of-the-art multiloss methods.

It improves training in multitask and transfer learning.

GradDrop reveals links between gradient stochasticity and optimal training.

Abstract

The vast majority of deep models use multiple gradient signals, typically corresponding to a sum of multiple loss terms, to update a shared set of trainable weights. However, these multiple updates can impede optimal training by pulling the model in conflicting directions. We present Gradient Sign Dropout (GradDrop), a probabilistic masking procedure which samples gradients at an activation layer based on their level of consistency. GradDrop is implemented as a simple deep layer that can be used in any deep net and synergizes with other gradient balancing approaches. We show that GradDrop outperforms the state-of-the-art multiloss methods within traditional multitask and transfer learning settings, and we discuss how GradDrop reveals links between optimal multiloss training and gradient stochasticity.