Research map / Architectures and efficiency / RNNs and normalization
Learning Frequency Domain Approximation for Binary Neural Networks
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 sine (Fourier series) gradient module to Dorefa-Net (ablation) | Dorefa-Net ResNet-20 · CIFAR-10 | accuracy (%): 84.44 -> 85.83 | Helped |
| add noise adaptation module on top of sine module (ablation) | FDA-BNN ResNet-20 · CIFAR-10 | accuracy (%): 85.83 -> 86.20 | Helped |
| add noise adaptation module to DSQ (ablation) | DSQ ResNet-20 · CIFAR-10 | accuracy (%): 84.11 -> 84.46 | Helped |
| add noise adaptation module to BNN+ (ablation) | BNN+ ResNet-20 · CIFAR-10 | accuracy (%): 84.59 -> 84.87 | Helped |
| add linear shortcut eta(x)=ax in noise module vs none (ablation) | FDA-BNN ResNet-20 · CIFAR-10 | accuracy (%): 85.27 -> 85.93 | Helped |
| use shortcut a*sin(x) instead of a*x (ablation) | FDA-BNN ResNet-20 · CIFAR-10 | accuracy (%): 85.93 -> 86.20 | Helped |
| increase number of Fourier series terms n beyond about 10 (ablation) | FDA-BNN · CIFAR-10 | accuracy (%) | No clear effect |
| gradually increase n starting from 9 rather than fixed n (ablation) | FDA-BNN · CIFAR-10 | accuracy (%) | Helped |
| replace STE gradient with Fourier series gradient in ReActNet | ReActNet ResNet-18 · ImageNet | top-1 accuracy (%): 65.5 -> 66.0 | Helped |
Closest papers by meaning
- Training Binary Neural Networks via Gaussian Variational Inference and Low-Rank Semidefinite Programming (NeurIPS 2024 · similarity 0.84)
- BiPer: Binary Neural Networks using a Periodic Function (CVPR 2024 · similarity 0.84)
- Addressing Spectral Bias of Deep Neural Networks by Multi-Grade Deep Learning (NeurIPS 2024 · similarity 0.83)
- Tuning the Frequencies: Robust Training for Sinusoidal Neural Networks (CVPR 2025 · similarity 0.82)
- Towards Accurate Binary Spiking Neural Networks: Learning with Adaptive Gradient Modulation Mechanism (AAAI 2025 · similarity 0.82)
- Fast and Slow Gradient Approximation for Binary Neural Network Optimization (AAAI 2025 · similarity 0.81)
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.