Pruning Binarized Neural Networks: A Dedicated Framework and Globally Weighted Algorithms

二值化神经网络剪枝：专用框架与全局加权算法

Abstract: Extreme compression of deep neural networks, up to full binarization, dramatically reduces memory footprint and arithmetic complexity, facilitating deployment on constrained edge hardware with field-programmable gate arrays (FPGAs) and microcontrollers. 摘要： 对深度神经网络进行极致压缩（直至完全二值化）可显著降低内存占用和算术复杂度，从而促进其在现场可编程门阵列（FPGA）和微控制器等受限边缘硬件上的部署。

Although combining binarization with pruning promises additional efficiency gains, existing pruning strategies are ill-suited to binarized representations and rarely translate into meaningful hardware savings. 尽管将二值化与剪枝相结合有望带来额外的效率提升，但现有的剪枝策略并不适用于二值化表示，且往往难以转化为实际的硬件性能收益。

We introduce a PyTorch-based, research-oriented framework that incorporates freezing and pruning mechanisms for designing and optimizing binarized neural networks. The framework enables rapid and reproducible evaluation of state-of-the-art approaches and the fast prototyping of new ones. 我们引入了一个基于 PyTorch 的研究型框架，该框架集成了用于设计和优化二值化神经网络的冻结（freezing）与剪枝机制。该框架支持对现有前沿方法进行快速且可复现的评估，并能快速构建新方法的原型。

Leveraging this framework, we propose a novel pruning method that accounts for the relative importance of learned parameters across abstraction levels. 利用该框架，我们提出了一种新的剪枝方法，该方法考虑了跨抽象层级学习参数的相对重要性。

Such a global weighting mechanism consistently achieves a superior trade-off between model accuracy and pruning rate, achieving a 70% pruning rate on VGG11 with constant accuracy, while state-of-the-art results reach only 41% in the binarized setting. 这种全局加权机制在模型精度和剪枝率之间始终能实现更优的权衡。在保持精度不变的情况下，该方法在 VGG11 上实现了 70% 的剪枝率，而目前二值化领域的前沿研究结果仅达到 41%。