BriefGPT.xyz
Oct, 2017
从动力系统观点设计的多层残差网络
Multi-level Residual Networks from Dynamical Systems View
HTML
PDF
Bo Chang, Lili Meng, Eldad Haber, Frederick Tung, David Begert
TL;DR
本文提出使用动力学系统视角来分析残差网络损伤特性，基于这些分析引入一种新的方法来加速残差网络的训练，并将该方法应用于图像分类等多个任务。
Abstract
Deep
residual networks
(ResNets) and their variants are widely used in many
computer vision
applications and
natural language processing
t
→