Research map / Architectures and efficiency / RNNs and normalization
Oscillatory Fourier Neural Network: A Compact and Efficient Architecture for Sequential Processing
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 |
|---|---|---|---|
| Add AC channels to the DC-only O-FNN (ablation) | O-FNN · IMDB sentiment analysis | learning performance | Helped |
| Increase base frequency above 2.0 (ablation) | O-FNN · IMDB sentiment analysis | test accuracy | Hurt |
| Increase total channels from 3 to 5 (ablation) | O-FNN · IMDB sentiment analysis | test accuracy | Helped |
| Add channels beyond 5 (ablation) | O-FNN · IMDB sentiment analysis | test accuracy | No clear effect |
| Use a 1D convolutional input layer with slicing window d=3 and stride s=1 instead of a fully connected input layer (ablation) | O-FNN · sequential processing tasks | accuracy | Helped |
| Add AC channels to the DC-only O-FNN (ablation) | O-FNN · permuted sMNIST | test accuracy | Helped |
| Increase channels beyond 3 (ablation) | O-FNN · permuted sMNIST | test accuracy | Hurt |
| Increase base frequency for longer sequences (ablation) | O-FNN · permuted sMNIST | test accuracy | Helped |
| Use more than four AC channels (ablation) | O-FNN · sequential learning tasks | accuracy | Hurt |
| Tune base frequency to a higher value for highly concentrated inputs in long sequences (ablation) | O-FNN · noise padded CIFAR-10 | test accuracy | Helped |
Closest papers by meaning
- Learning Long Term Dependencies via Fourier Recurrent Units (ICML 2018 · similarity 0.87)
- Coupled Oscillatory Recurrent Neural Network (coRNN): An accurate and (gradient) stable architecture for learning long time dependencies (ICLR 2021 · similarity 0.87)
- Fast-Slow Recurrent Neural Networks (NeurIPS 2017 · similarity 0.86)
- Efficient recurrent architectures through activity sparsity and sparse back-propagation through time (ICLR 2023 · similarity 0.86)
- Parallelizing non-linear sequential models over the sequence length (ICLR 2024 · similarity 0.86)
- FFN Fusion: Rethinking Sequential Computation in Large Language Models (NeurIPS 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.