ResidualNetWorks总结
ResNets:即残差网络,能帮助解决由于网络结构过于deep而产生的梯度消失问题。
Two main types of blocks are used in a ResNet, depending mainly on whether the input/output dimensions are same or different:
The identity block is the standard block used in ResNets, and corresponds to the case where the input activation(say a[l]) has the same dimension as the output activation(say a[l+2]), here is an alternative diagram showing the individual steps:
Figure 1 : Identity block . Skip connection "skip over" 2 layers.
Figure 2 : Identity block . Skip connection "skip over" 3 layers.
Figure 3 : Convolutiional block .
上述两种方式,代码实现起来也比较简单,可以参考deeplearning作业(这里用keras实现):https://github.com/NumberFairy/deep_learning_answer/blob/master/test4_2_residualNetworks/residualNetworks/index.py