Learning Longer-term Dependencies in RNNs with Auxiliary Losses

TL;DR

This paper introduces an auxiliary loss method to enhance RNNs' ability to learn long-term dependencies, enabling training on very long sequences more efficiently and effectively.

Contribution

The paper proposes a simple auxiliary loss approach that improves long-term dependency learning in RNNs and scales to very long sequences, outperforming existing methods.

Findings

Effective on sequences up to 16,000 steps

Outperforms competitive baselines including Transformers

Improves optimization and regularization

Abstract

Despite recent advances in training recurrent neural networks (RNNs), capturing long-term dependencies in sequences remains a fundamental challenge. Most approaches use backpropagation through time (BPTT), which is difficult to scale to very long sequences. This paper proposes a simple method that improves the ability to capture long term dependencies in RNNs by adding an unsupervised auxiliary loss to the original objective. This auxiliary loss forces RNNs to either reconstruct previous events or predict next events in a sequence, making truncated backpropagation feasible for long sequences and also improving full BPTT. We evaluate our method on a variety of settings, including pixel-by-pixel image classification with sequence lengths up to 16\,000, and a real document classification benchmark. Our results highlight good performance and resource efficiency of this approach over competitive baselines, including other recurrent models and a comparable sized Transformer. Further analyses reveal beneficial effects of the auxiliary loss on optimization and regularization, as well as extreme cases where there is little to no backpropagation.