Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers

TL;DR

This paper introduces AFNO, an efficient Fourier-based token mixer for vision transformers, which reduces computational complexity and improves performance on high-resolution image segmentation tasks.

Contribution

The paper proposes AFNO, a novel Fourier neural operator-based token mixer with architectural modifications for efficient high-resolution vision transformer processing.

Findings

AFNO outperforms self-attention in few-shot segmentation accuracy.

AFNO handles sequence sizes of 65k efficiently in Cityscapes segmentation.

AFNO achieves quasi-linear complexity and linear memory in sequence size.

Abstract

Vision transformers have delivered tremendous success in representation learning. This is primarily due to effective token mixing through self attention. However, this scales quadratically with the number of pixels, which becomes infeasible for high-resolution inputs. To cope with this challenge, we propose Adaptive Fourier Neural Operator (AFNO) as an efficient token mixer that learns to mix in the Fourier domain. AFNO is based on a principled foundation of operator learning which allows us to frame token mixing as a continuous global convolution without any dependence on the input resolution. This principle was previously used to design FNO, which solves global convolution efficiently in the Fourier domain and has shown promise in learning challenging PDEs. To handle challenges in visual representation learning such as discontinuities in images and high resolution inputs, we propose principled architectural modifications to FNO which results in memory and computational efficiency. This includes imposing a block-diagonal structure on the channel mixing weights, adaptively sharing weights across tokens, and sparsifying the frequency modes via soft-thresholding and shrinkage. The resulting model is highly parallel with a quasi-linear complexity and has linear memory in the sequence size. AFNO outperforms self-attention mechanisms for few-shot segmentation in terms of both efficiency and accuracy. For Cityscapes segmentation with the Segformer-B3 backbone, AFNO can handle a sequence size of 65k and outperforms other efficient self-attention mechanisms.