Research map / Architectures and efficiency / Quantization
Equal Bits: Enforcing Equally Distributed Binary Network Weights
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 |
|---|---|---|---|
| Enforce equal positive and negative weight ratios with optimal-transport binarization | Binary networks · CIFAR-10, Conv2/4/6/8 | accuracy: 77.86± 0.69 -> 79.25± 0.28; 86.49± 0.24 -> 87.68± 0.32; 88.51± 0.35 -> 89.92± 0.19; 89.17±0.26 -> 90.40±0.17 | Helped |
| Enforce equal positive and negative weight ratios with optimal-transport binarization | Binary ResNet-18 · ImageNet | Top-1 accuracy: 59.98 -> 60.40 | Helped |
| Enforce equal positive and negative weight ratios with optimal-transport binarization | Binary ResNet-34 · ImageNet | Top-1 accuracy: 63.1 -> 64.17 | Helped |
| Vary the positive-to-negative binary-weight prior, using equal ratios (ablation) | Binary Conv2 · CIFAR-10 and CIFAR-100 | accuracy | Helped |
| Use optimal-transport equal-ratio binarization instead of the sign function | Conv2 on CIFAR-10 across repeated training runs · CIFAR-10 | accuracy | Helped |
| Apply equal-ratio optimal-transport binarization to remaining unpruned weights | Pruned binary networks · CIFAR-10 Conv2/4/6/8 and CIFAR-100 ResNet-18 | test accuracy | Helped |
Closest papers by meaning
- BiBench: Benchmarking and Analyzing Network Binarization (ICML 2023 · similarity 0.87)
- Forward and Backward Information Retention for Accurate Binary Neural Networks (CVPR 2020 · similarity 0.86)
- QBB: Quantization with Binary Bases for LLMs (NeurIPS 2024 · similarity 0.84)
- Understanding the Unfairness in Network Quantization (ICML 2025 · similarity 0.84)
- BinaryDM: Accurate Weight Binarization for Efficient Diffusion Models (ICLR 2025 · similarity 0.84)
- Training Binary Neural Networks via Gaussian Variational Inference and Low-Rank Semidefinite Programming (NeurIPS 2024 · similarity 0.82)
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.