Identity Mappings in Deep Residual Networks

16 Mar 2016Kaiming HeXiangyu ZhangShaoqing RenJian Sun

Deep residual networks have emerged as a family of extremely deep architectures showing compelling accuracy and nice convergence behaviors. In this paper, we analyze the propagation formulations behind the residual building blocks, which suggest that the forward and backward signals can be directly propagated from one block to any other block, when using identity mappings as the skip connections and after-addition activation... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Image Classification CIFAR-10 ResNet-1001 Percentage correct 95.4 # 16
Image Classification CIFAR-10 ResNet-1001 Percentage error 4.62 # 11
Image Classification CIFAR-100 ResNet-1001 Percentage correct 77.3 # 12
Image Classification CIFAR-100 ResNet-1001 Percentage error 22.71 # 6