【转载】 xavier,kaiming初始化中的fan_in,fan_out在卷积神经网络是什么意思
作者:互联网
原文地址:
https://www.cnblogs.com/liuzhan709/p/10092679.html
=======================================================
xavier初始化出自论文Understanding the difficulty of training deep feedforward neural network, 论文讨论的是全连接神经网络,fan_in指第i层神经元个数,fan_out指第i+1层神经元个数,但是我们的卷积神经网路是局部连接的,此时的fan_in,fan_out是什么意思呢。
在pytorch中:
fan_in指 kernel_height x kernel_width x in_channel 。
fan_out指 kernel_height x kernel_width x out_channel 。
从局部连接的过程来看似乎并不十分合理,卷积神经网络的局部连接在感受野内仍然是全连接。fan_in=kh x kw x in_channel没什么疑问,但是fan_out应该等于out_channel更合理啊。待解答。
code,来自pytorch实现
def _calculate_fan_in_and_fan_out(tensor): dimensions = tensor.ndimension() if dimensions < 2: raise ValueError("Fan in and fan out can not be computed for tensor with fewer than 2 dimensions") if dimensions == 2: # Linear fan_in = tensor.size(1) fan_out = tensor.size(0) else: num_input_fmaps = tensor.size(1) num_output_fmaps = tensor.size(0) receptive_field_size = 1 if tensor.dim() > 2: receptive_field_size = tensor[0][0].numel() fan_in = num_input_fmaps * receptive_field_size fan_out = num_output_fmaps * receptive_field_size return fan_in, fan_out
==============================================
标签:kernel,tensor,kaiming,dimensions,fan,xavier,out,size 来源: https://www.cnblogs.com/devilmaycry812839668/p/16074429.html