Research map / Architectures and efficiency / Quantization
Forward and Backward Information Retention for Accurate Binary Neural Networks
Check 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 |
|---|---|---|---|
| Use IR-Net instead of RAD | ResNet-18 · CIFAR-10 | accuracy: 90.5 -> 91.5 | Helped |
| Use IR-Net (Bi-Real structure) instead of DSQ | ResNet-20 · CIFAR-10 | accuracy: 84.1 -> 86.5 | Helped |
| Use IR-Net instead of BNN | VGG-Small · CIFAR-10 | accuracy: 89.9 -> 90.4 | Helped |
| Use IR-Net (Bi-Real structure) instead of XNOR++ | ResNet-18 · ImageNet | Top-1 accuracy: 57.1 -> 58.1 | Helped |
| Use IR-Net instead of BWHN | ResNet-18 · ImageNet | Top-1 accuracy: 64.3 -> 66.5 | Helped |
| Use IR-Net (Bi-Real structure) instead of Bi-Real | ResNet-34 · ImageNet | Top-1 accuracy: 62.2 -> 62.9 | Helped |
| Add weight standardization to Libra-PB (ablation) | IR-Net on ResNet-20 · CIFAR-10 | accuracy: 84.3 -> 84.9 | Helped |
| Add bit-shift scales to Libra-PB (ablation) | IR-Net on ResNet-20 · CIFAR-10 | accuracy: 84.6 -> 84.9 | Helped |
| Add EDE to Libra-PB (ablation) | IR-Net on ResNet-20 · CIFAR-10 | accuracy: 84.9 -> 86.5 | Helped |
| Add Libra-PB to EDE (ablation) | IR-Net on ResNet-20 · CIFAR-10 | accuracy: 85.2 -> 86.5 | Helped |
| Use Libra-PB alone instead of binary baseline (ablation) | ResNet-20 · CIFAR-10 | accuracy: 83.8 -> 84.9 | Helped |
| Use EDE alone instead of binary baseline (ablation) | ResNet-20 · CIFAR-10 | accuracy: 83.8 -> 85.2 | Helped |
Closest papers by meaning
- Understanding Neural Network Binarization with Forward and Backward Proximal Quantizers (NeurIPS 2023 · similarity 0.89)
- BiBench: Benchmarking and Analyzing Network Binarization (ICML 2023 · similarity 0.88)
- Sparsity-Inducing Binarized Neural Networks (AAAI 2020 · similarity 0.88)
- Fast and Slow Gradient Approximation for Binary Neural Network Optimization (AAAI 2025 · similarity 0.86)
- Training Binary Neural Networks via Gaussian Variational Inference and Low-Rank Semidefinite Programming (NeurIPS 2024 · similarity 0.86)
- Layerwise Progressive Freezing: A Training Scaffold for Depth-Scalable Binary Networks (NeurIPS 2026 · 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.