Research map / Architectures and efficiency / RNNs and normalization
Learning Long Term Dependencies via Fourier Recurrent Units
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the SRU with FRU₆₀,₁₀ | recurrent model · Permuted MNIST | testing accuracy: 92.21% -> 96.93% | Helped |
| replace the LSTM with FRU₆₀,₁₀ | recurrent model · Permuted MNIST | testing accuracy: 90.26% -> 96.93% | Helped |
| replace the RNN with FRU₆₀,₁₀ | recurrent model · Permuted MNIST | testing accuracy: 87.46% -> 96.93% | Helped |
| replace the SRU with FRU₅,₁₀ | recurrent model · IMDB | testing accuracy: 86.40% -> 86.71% | Helped |
| replace the LSTM with FRU₅,₁₀ | recurrent model · IMDB | testing accuracy: 83.64% -> 86.71% | Helped |
| use only frequency 0 instead of five frequencies (ablation) | FRU on IMDB · IMDB | testing accuracy: 86.71% -> 86.44% | No clear effect |
| use 40 frequencies instead of 60 (ablation) | FRU on Pixel-MNIST · Pixel-MNIST | testing accuracy: 97.61% -> 96.88% | Hurt |
| replace the LSTM with FRU₆₀,₁₀ | recurrent model · Pixel-MNIST | testing accuracy: 98.17% -> 97.61% | Hurt |
| replace the SRU with FRU₆₀,₁₀ | recurrent model · Pixel-MNIST | testing accuracy: 96.20% -> 97.61% | Helped |
Closest papers by meaning
- Towards Non-Saturating Recurrent Units for Modelling Long-Term Dependencies (AAAI 2019 · similarity 0.89)
- Oscillatory Fourier Neural Network: A Compact and Efficient Architecture for Sequential Processing (AAAI 2022 · similarity 0.87)
- Long Expressive Memory for Sequence Modeling (ICLR 2022 · similarity 0.87)
- Recurrent neural networks: vanishing and exploding gradients are not the end of the story (NeurIPS 2024 · similarity 0.86)
- Rethinking Fourier Transform from A Basis Functions Perspective for Long-term Time Series Forecasting (NeurIPS 2024 · similarity 0.85)
- Plastic Learning with Deep Fourier Features (ICLR 2025 · similarity 0.85)
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.