Just Pick a Sign: Optimizing Deep Multitask Models with Gradient Sign Dropout
This addresses the challenge of training deep models with multiple loss terms for researchers and practitioners in machine learning, offering an incremental improvement over existing gradient balancing approaches.
The paper tackles the problem of conflicting gradient signals in deep multitask models, which impede optimal training, by introducing Gradient Sign Dropout (GradDrop), a probabilistic masking procedure that samples gradients based on consistency. The result shows that GradDrop outperforms state-of-the-art multiloss methods in multitask and transfer learning settings.
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.