用率约束量化+熵编码，让神经网络压缩后体积减少20%-40%。
Reducing Storage of Pretrained Neural Networks by Rate-Constrained Quantization and Entropy Coding
- 引入二次率估计改进层间损失，实现更精准的压缩控制。
- 在保持原性能前提下，比特率降低20%-40%，优于NNCodec。
- 解码速度快，支持任意量化网格，适合嵌入式设备部署。
神经网络规模持续增长，对嵌入式传感器等资源受限设备带来挑战。压缩算法可缓解此问题，前提是模型性能接近原始水平。本文提出一种新型后训练压缩框架，结合率感知量化与熵编码：(1) 在经典层间损失基础上引入二次率估计；(2) 基于最优剪枝思想（OBS）提供该目标的局部精确解。方法支持任意量化网格，解码极快。在多个视觉网络上验证，相同性能下比特率较主流压缩算法NNCodec降低20%-40%。代码已开源：https://github.com/Conzel/cerwu。
原文摘要 · Abstract (English)
The ever-growing size of neural networks poses serious challenges on resource-constrained devices, such as embedded sensors. Compression algorithms that reduce their size can mitigate these problems, provided that model performance stays close to the original. We propose a novel post-training compression framework that combines rate-aware quantization with entropy coding by (1) extending the well-known layer-wise loss by a quadratic rate estimation, and (2) providing locally exact solutions to this modified objective following the Optimal Brain Surgeon (OBS) method. Our method allows for very fast decoding and is compatible with arbitrary quantization grids. We verify our results empirically by testing on various computer-vision networks, achieving a 20-40\% decrease in bit rate at the same performance as the popular compression algorithm NNCodec. Our code is available at https://github.com/Conzel/cerwu.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。