Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers

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.

References in corpus (17)

• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• MLP-Mixer: An all-MLP Architecture for Vision
• Fourier Neural Operator for Parametric Partial Differential Equations
• SegFormer: Simple and Efficient Design for Semantic Segmentation with Transformers
• Linformer: Self-Attention with Linear Complexity
• Generating Long Sequences with Sparse Transformers
• Axial Attention in Multidimensional Transformers
• Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention
• Rethinking Attention with Performers
• AS-MLP: An Axial Shifted MLP Architecture for Vision
• Sparse Sinkhorn Attention
• Multiwavelet-based Operator Learning for Differential Equations
• Choose a Transformer: Fourier or Galerkin
• Long-Short Transformer: Efficient Transformers for Language and Vision
• Pay Attention to MLPs
• Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention
• Global Filter Networks for Image Classification

Cited by in corpus (3)

• Long-term predictions of turbulence by implicit U-Net enhanced Fourier neural operator
• FengWu: Pushing the Skillful Global Medium-range Weather Forecast beyond 10 Days Lead
• Accelerating Phase Field Simulations Through a Hybrid Adaptive Fourier Neural Operator with U-Net Backbone