AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Choose a Transformer: Fourier or Galerkin

Shuhao Cao

NeurIPS 2021 · Architectures and efficiency › Attention and transformers · approach: Attention head and block redesign · 57 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Replace softmax attention with Galerkin-type attentionoperator learners for Burgers and Darcy solution operators · Burgers, n=8192relative error (× 10^-3): 4.151 -> 1.025Improved
Use an attention-based operator learner instead of FNODarcy flow solution operator · Darcy flow, n_f,n_c=141,43relative error (× 10^-2): 1.419 -> 0.838Improved
Use attention-based operator learners instead of FNOinverse Darcy coefficient identification · Inverse Darcy, n_f,n_c=141,36, ε=0.1relative error (× 10^-2): 15.12 -> 6.659Improved
Use the Galerkin projection-type layer normalization rather than regular layer normalization (ablation)GT on unnormalized Burgers data · Burgers, n=8192relative error (× 10^-3): 2.747 -> 1.025Improved
Use the Galerkin projection-type layer normalization rather than regular layer normalization (ablation)GT on normalized Darcy data · Darcy flow, n_f,n_c=141,43relative error (× 10^-2): 0.894 -> 0.839No clear effect
Add diagonal initialization with η=δ=10^-2 (ablation)GT with the new normalization on Burgers · Burgers, n=8192relative error (× 10^-3): 9.129 -> 1.025Improved
Add diagonal initialization with η=δ=10^-2 (ablation)GT with regular normalization on Burgers · Burgers, n=512relative error (× 10^-3): 200.4 -> 15.14Improved
Add attention dropout ζ=0.1 (ablation)GT with regular normalization on Burgers · Burgers, n=8192relative error (× 10^-3): 217.5 -> 207.0Improved
Remove the softmax normalizationattention-based operator learners · training efficiencyGFLOP per BP: 1393 -> 275Improved
Use Galerkin attention instead of linear softmax attentionencoder-only profiling, n=8192, d=128, l=10 · training speediterations per second: 10.98 -> 19.20Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy