WavSpA: Wavelet Space Attention for Boosting Transformers' Long Sequence Learning Ability
This addresses the bottleneck of long sequence processing in Transformers, offering a domain-specific improvement for tasks requiring extended context.
The paper tackles the problem of improving Transformers' long sequence learning ability by proposing Wavelet Space Attention (WavSpA), which replaces standard attention with attention learning in a wavelet coefficient space, resulting in consistent performance gains on the Long Range Arena and enhanced reasoning extrapolation on the LEGO task.
Transformer and its variants are fundamental neural architectures in deep learning. Recent works show that learning attention in the Fourier space can improve the long sequence learning capability of Transformers. We argue that wavelet transform shall be a better choice because it captures both position and frequency information with linear time complexity. Therefore, in this paper, we systematically study the synergy between wavelet transform and Transformers. We propose Wavelet Space Attention (WavSpA) that facilitates attention learning in a learnable wavelet coefficient space which replaces the attention in Transformers by (1) applying forward wavelet transform to project the input sequences to multi-resolution bases, (2) conducting attention learning in the wavelet coefficient space, and (3) reconstructing the representation in input space via backward wavelet transform. Extensive experiments on the Long Range Arena demonstrate that learning attention in the wavelet space using either fixed or adaptive wavelets can consistently improve Transformer's performance and also significantly outperform learning in Fourier space. We further show our method can enhance Transformer's reasoning extrapolation capability over distance on the LEGO chain-of-reasoning task.