TensorFlow批处理规范化实现之间有什么区别？
问题描述：
TensorFlow似乎实现了至少三个版本的批量规范化：
TensorFlow seems to implement at least 3 versions of batch normalization:
tf.nn.batch_normalizationtf.layers.batch_normalizationtf.contrib.layers.batch_norm
所有这些有不同的论据和文档。
These all have different arguments and documentation.
这些和我应该使用哪一个？
What is the difference between these, and which one should I use?
答
它们实际上是非常不同的。
They are actually very different.
-
nn.batch_normalization执行基本操作（即简单的标准化） -
layers.batch_norma化是一个批处理层，即它负责设置可训练的参数等。最终，它是nn.batch_normalization 的包装。 code>。除非您想自己设置变量等，否则这是您要使用的一种。
-
nn.batch_normalizationperforms the basic operation (i.e. a simple normalization) -
layers.batch_normalizationis a batchnorm "layer", i.e. it takes care of setting up the trainable parameters etc. At the end of the day, it is a wrapper aroundnn.batch_normalization. Chances are this is the one you want to use, unless you want to take care of setting up variables etc. yourself.
例如， nn.conv2d 和 layers.conv2d 之间的差异。
对于 contrib 版本，我不能肯定地说，但是在我看来，它像是一个实验版本，带有一些附加参数，在常规 层一层。
As for the contrib version, I can't say for sure, but it seems to me like an experimental version with some extra parameters not available in the "regular" layers one.