BriefGPT.xyz
Dec, 2017
神经网络的量化和训练，用于高效的整数运算推理
Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference
HTML
PDF
Benoit Jacob, Skirmantas Kligys, Bo Chen, Menglong Zhu, Matthew Tang...
TL;DR
本文提出了一种量化方案，通过整数运算进行推断，以提高在移动设备上的效率，并设计了一种训练程序来维护量化后的模型精度。该方案在MobileNets模型中展现了显著的改进，在ImageNet分类和COCO检测等任务上获得了良好的结果。
Abstract
The rising popularity of intelligent
mobile devices
and the daunting computational cost of deep learning-based models call for efficient and accurate
on-device inference
schemes. We propose a
→