Relay Backpropagation for Effective Learning of Deep Convolutional Neural Networks

被引：302

作者：

Shen, Li ^{[1
,2
]}

Lin, Zhouchen ^{[3
,4
]}

Huang, Qingming ^{[1
]}

机构：

[1] Univ Chinese Acad Sci, Beijing, Peoples R China

[2] Univ Oxford, Oxford OX1 2JD, England

[3] Peking Univ, Sch EECS, MOE, Key Lab Machine Percept, Beijing, Peoples R China

[4] Shanghai Jiao Tong Univ, Cooperat Medianet Innovat Ctr, Shanghai, Peoples R China

来源：

COMPUTER VISION - ECCV 2016, PT VII | 2016年 / 9911卷

关键词：

Relay Backpropagation; Convolutional neural networks; Large scale image classification;

D O I：

10.1007/978-3-319-46478-7_29

中图分类号：

TP18 [人工智能理论];

学科分类号：

081104 ; 0812 ; 0835 ; 1405 ;

摘要：

Learning deeper convolutional neural networks has become a tendency in recent years. However, many empirical evidences suggest that performance improvement cannot be attained by simply stacking more layers. In this paper, we consider the issue from an information theoretical perspective, and propose a novel method Relay Backpropagation, which encourages the propagation of effective information through the network in training stage. By virtue of the method, we achieved the first place in ILSVRC 2015 Scene Classification Challenge. Extensive experiments on two large scale challenging datasets demonstrate the effectiveness of our method is not restricted to a specific dataset or network architecture.

引用

页码：467 / 482

页数：16