Improving the Gating Mechanism of Recurrent Neural Networks
This addresses a key bottleneck in training recurrent models for tasks with long-term dependencies, though it is incremental as it builds on existing gating mechanisms.
The paper tackled the problem of gating mechanisms in recurrent neural networks operating in saturation, which hinders gradient-based learning, by introducing two easy-to-implement modifications that improve learnability without extra hyperparameters. The result showed robust performance improvements on tasks like synthetic memorization, sequential image classification, language modeling, and reinforcement learning, especially for long-term dependencies.
Gating mechanisms are widely used in neural network models, where they allow gradients to backpropagate more easily through depth or time. However, their saturation property introduces problems of its own. For example, in recurrent models these gates need to have outputs near 1 to propagate information over long time-delays, which requires them to operate in their saturation regime and hinders gradient-based learning of the gate mechanism. We address this problem by deriving two synergistic modifications to the standard gating mechanism that are easy to implement, introduce no additional hyperparameters, and improve learnability of the gates when they are close to saturation. We show how these changes are related to and improve on alternative recently proposed gating mechanisms such as chrono initialization and Ordered Neurons. Empirically, our simple gating mechanisms robustly improve the performance of recurrent models on a range of applications, including synthetic memorization tasks, sequential image classification, language modeling, and reinforcement learning, particularly when long-term dependencies are involved.