Variable Computation in Recurrent Neural Networks

Abstract

Recurrent neural networks (RNNs) have been used extensively and with increasing success to model various types of sequential data. Much of this progress has been achieved through devising recurrent units and architectures with the flexibility to capture complex statistics in the data, such as long range dependency or localized attention phenomena. However, while many sequential data (such as video, speech or language) can have highly variable information flow, most recurrent models still consume input features at a constant rate and perform a constant number of computations per time step, which can be detrimental to both speed and model capacity. In this paper, we explore a modification to existing recurrent units which allows them to learn to vary the amount of computation they perform at each step, without prior knowledge of the sequence's time structure. We show experimentally that not only do our models require fewer operations, they also lead to better performance overall on evaluation tasks.

Cited by in corpus (12)

• Neural Ordinary Differential Equations
• Recurrent Independent Mechanisms
• Neural Speed Reading via Skim-RNN
• Dynamic Neural Networks: A Survey
• Skip RNN: Learning to Skip State Updates in Recurrent Neural Networks
• Dynamic Computational Time for Visual Attention
• Surprisal-Triggered Conditional Computation with Neural Networks
• On the Transformer Growth for Progressive BERT Training
• Depth-Adaptive Graph Recurrent Network for Text Classification
• Learning Simpler Language Models with the Differential State Framework
• A Survey on Green Deep Learning
• Adaptive Attention Span in Transformers