Research map / Architectures and efficiency / Attention and transformers
FourierFormer: Transformer Meets Generalized Fourier Integral Theorem
Tan Minh Nguyen, Minh Pham, Tam Minh Nguyen, Khai Nguyen, Stanley Osher, Nhat Ho
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace dot-product attention with Fourier attention (small) | Transformer LM · WikiText-103 test | perplexity: 34.29 -> 32.85 | Improved |
| replace dot-product attention with Fourier attention (medium) | Transformer LM · WikiText-103 test | perplexity: 29.60 -> 28.01 | Improved |
| increase model size from small to medium | FourierFormer vs baseline · WikiText-103 valid | perplexity gain: 1.29 -> 1.39 | Improved |
| replace dot-product attention with Fourier attention | DeiT · ImageNet | top-1 accuracy: 72.23 -> 73.25 | Improved |
| replace dot-product attention with Fourier attention | DeiT · ImageNet | top-5 accuracy: 91.13 -> 91.66 | Improved |
| replace softmax attention with Fourier attention | Transformer · UEA average | accuracy: 72.07 -> 73.17 | Improved |
| replace softmax attention with Fourier attention | Transformer · UEA PEMS-SF | accuracy: 82.66 -> 86.70 | Improved |
| replace softmax attention with Fourier attention | Transformer · UEA SpokenArabicDigits | accuracy: 99.33 -> 99.00 | Hurt |
| replace softmax attention with Fourier attention | Transformer · UEA Handwriting | accuracy: 32.08 -> 31.68 | Hurt |
| replace softmax attention with Fourier attention | Decision Transformer · D4RL average | reward: 74.39 -> 75.69 | Improved |
| replace softmax attention with Fourier attention | Decision Transformer · D4RL medium-replay Walker | reward: 66.11 -> 63.19 | Hurt |
| replace softmax attention with Fourier attention | Transformer · IWSLT14 De-En | BLEU: 34.42 -> 34.68 | Improved |
Closest papers by meaning
- Sparse Attention as Compact Kernel Regression (NeurIPS 2026 · similarity 0.86)
- PolySketchFormer: Fast Transformers via Sketching Polynomial Kernels (ICML 2024 · similarity 0.85)
- DiJiang: Efficient Large Language Models through Compact Kernelization (ICML 2024 · similarity 0.85)
- RelFlexformer: Efficient Attention Transformers for Integrable Relative Positional Encodings (NeurIPS 2026 · similarity 0.84)
- Fourier Head: Helping Large Language Models Learn Complex Probability Distributions (ICLR 2025 · similarity 0.84)
- Theory, Analysis, and Best Practices for Sigmoid Self-Attention (ICLR 2025 · similarity 0.84)
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.
