Neural network 在Apple提供的MetalImageRecognition示例代码中，如何对权重和偏差参数执行批量标准化？
自述文件提供了以下评论：Neural network 在Apple提供的MetalImageRecognition示例代码中，如何对权重和偏差参数执行批量标准化？,neural-network,ios10,conv-neural-network,metal,Neural Network,Ios10,Conv Neural Network,Metal,自述文件提供了以下评论： /* The weights for this particular network were batch normalized but for inference we may use : w = gamma / √(s + 0.001), b = ß - ( A * m ) s: variance m: mean gamma : gamma ß: beta w: weights of a feature channel b: bias of
/*
The weights for this particular network were batch normalized but for inference we may use :
w = gamma / √(s + 0.001), b = ß - ( A * m )
s: variance
m: mean
gamma : gamma
ß: beta
w: weights of a feature channel
b: bias of a feature channel
for every feature channel separately to get the corresponding weights and bias
*/
谢谢。我创建了一个回购协议，将初始v3批次标准化权重转换为MPSCNN卷积所需的非标准化权重 在Sergey Ioffe和Christian Szegedy（）撰写的论文“批量规范化：通过减少内部协变量移位来加速深层网络训练”中，我们可以使用算法2，输出，步骤11推导：
[Weight = \frac{\gamma}{\sqrt{Var[x]+0.001}} * Weight_{BN}](http://mathurl.com/z7snq3z.png)
[Bias = \beta - (\frac{\gamma}{\sqrt{Var[x]+0.001}}) * E[x]](http://mathurl.com/zo4shhf.png)