arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.12422cs.LGcs.AI
收藏

基于标准偏差的深度神经网络Quantization

Standard Deviation-Based Quantization for Deep Neural Networks

McGill University(麦吉尔大学)

Amir Ardakani, Arash Ardakani, Brett Meyer, James J. Clark, Warren J. Gross

摘要:

深度神经网络的量化是一种很有前途的方法,可以降低推理成本,使在资源受限的设备上运行深度网络变得可行。受现有方法的启发,我们提出了一个新的框架,使用网络的权重和激活分布的知识来学习量化区间(离散值),即,标准偏差此外,我们提出了一种新的基2对数量化方案,将权重转换为2的幂次离散值。我们提出的方案,使我们能够取代资源饥饿的高精度乘法器与简单的移位加操作。根据我们的评估,与全精度模型相比,我们的方法优于CIFAR 10和ImageNet数据集上的现有工作,甚至在3位权重和激活下实现了更好的准确性性能。此外,我们的方案同时修剪网络的参数,并允许我们灵活地调整修剪率在量化过程中。

英文摘要:

Quantization of deep neural networks is a promising approach that reduces the inference cost, making it feasible to run deep networks on resource-restricted devices. Inspired by existing methods, we propose a new framework to learn the quantization intervals (discrete values) using the knowledge of the network's weight and activation distributions, i.e., standard deviation. Furthermore, we propose a novel base-2 logarithmic quantization scheme to quantize weights to power-of-two discrete values. Our proposed scheme allows us to replace resource-hungry high-precision multipliers with simple shift-add operations. According to our evaluations, our method outperforms existing work on CIFAR10 and ImageNet datasets and even achieves better accuracy performance with 3-bit weights and activations when compared to the full-precision models. Moreover, our scheme simultaneously prunes the network's parameters and allows us to flexibly adjust the pruning ratio during the quantization process.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑