Learning Long Term Dependencies via Fourier Recurrent Units
This addresses the vanishing/exploding gradient problem in recurrent neural networks, offering a potential improvement for sequence modeling tasks, though it appears incremental as it builds on existing recurrent architectures.
The paper tackles the challenge of training recurrent neural networks for tasks with long-term dependencies by proposing the Fourier Recurrent Unit (FRU), which stabilizes gradients and provides stronger expressive power, outperforming other architectures with fewer parameters.
It is a known fact that training recurrent neural networks for tasks that have long term dependencies is challenging. One of the main reasons is the vanishing or exploding gradient problem, which prevents gradient information from propagating to early layers. In this paper we propose a simple recurrent architecture, the Fourier Recurrent Unit (FRU), that stabilizes the gradients that arise in its training while giving us stronger expressive power. Specifically, FRU summarizes the hidden states $h^{(t)}$ along the temporal dimension with Fourier basis functions. This allows gradients to easily reach any layer due to FRU's residual learning structure and the global support of trigonometric functions. We show that FRU has gradient lower and upper bounds independent of temporal dimension. We also show the strong expressivity of sparse Fourier basis, from which FRU obtains its strong expressive power. Our experimental study also demonstrates that with fewer parameters the proposed architecture outperforms other recurrent architectures on many tasks.