Google Colab中使用tensorflow.keras.Normalization报错求助

用户在Google Colab中运行以下代码时出现Reshape错误：

import tensorflow as tf
import pandas as pd
from tensorflow.keras.layers import Normalization

normalizer = Normalization(mean=5, variance=4)
x_normalized = tf.constant([[3,4,5,6,7],
                            [4,5,6,7,8]])
normalizer(x_normalized)

报错信息：

InvalidArgumentError: {{function_node __wrapped__Reshape_device_/job:localhost/replica:0/task:0/device:CPU:0}} Input to reshape is a tensor with 1 values, but the requested shape has 5 [Op:Reshape]

错误原因 #

你传入的mean和variance是单个标量值，但输入张量x_normalized是2样本×5特征的结构（每个样本包含5个特征）。Normalization层要求mean和variance的维度必须与输入的特征维度一致（这里需要长度为5的数组），否则层在尝试对齐维度时会触发Reshape错误。

两种解决方法 #

• 手动匹配特征维度的均值方差
  如果你明确知道每个特征的均值和方差，传入与特征数一致的数组即可：

  import tensorflow as tf
  from tensorflow.keras.layers import Normalization
  
  # 为5个特征分别指定均值和方差
  normalizer = Normalization(mean=[5,5,5,5,5], variance=[4,4,4,4,4])
  x_normalized = tf.constant([[3,4,5,6,7],
                            [4,5,6,7,8]])
  print(normalizer(x_normalized))
  

• 使用adapt方法自动计算统计量（更推荐）
  如果没有预设的均值方差，可让Normalization层通过输入数据自动计算统计量：

  import tensorflow as tf
  from tensorflow.keras.layers import Normalization
  
  normalizer = Normalization()
  x_normalized = tf.constant([[3,4,5,6,7],
                            [4,5,6,7,8]])
  # 用输入数据适配层，自动计算均值和方差
  normalizer.adapt(x_normalized)
  print(normalizer(x_normalized))
  

内容的提问来源于stack exchange，提问作者bbax4200