BriefGPT.xyz
Sep, 2018
可学习正则化的稀疏低精度神经网络学习
Learning Low Precision Deep Neural Networks through Regularization
HTML
PDF
Yoojin Choi, Mostafa El-Khamy, Jungwon Lee
TL;DR
本文提出了一种使用低精度权重和操作的DNN学习方法,利用可学习的正则化系数来加强高精度权重收敛到量化值的能力,并研究了如何通过权重剪枝、量化和熵编码来建立低精度DNN压缩管道,实验结果显示该方法可以在ImageNet分类和图像超分辨率网络的任务中取得与准确性相对可接受的最新压缩比。
Abstract
We consider the quantization of deep neural networks (
dnns
) to produce
low-precision
models for efficient inference of fixed-point operations. Compared to previous approaches to training quantized
→