TensorFlow 中的 tf.transpose()函数

tf.transpose()是 TensorFlow 中提供的函数。该函数用于转置输入张量。

Syntax: tf.transpose(input_tensor, perm, conjugate)

Parameters:
input_tensor: as the name suggests it is the tensor which is to be transposed.
Type: Tensor

perm: This parameters specifies the permutation according to which the input_tensor is to be transposed.
Type: Vector

conjugate: This parameters is set to True if the input_tensor is of type complex.
Type: Boolean

示例 1：

import tensorflow as geek
  
  
x = geek.constant([[1, 2, 3, 4],
                   [5, 6, 7, 8]])
transposed_tensor = geek.transpose(x)

输出 ：

array([[1, 5],
       [2, 6],
       [3, 7],
       [4, 8]])

示例 2：使用 perm 参数：

当此参数通过时，张量沿给定轴转置。简单来说，它定义了转置张量的输出形状。

import tensorflow as geek
  
  
x = geek.constant([[[ 1, 2, 3],
                    [ 4, 5, 6]],
                   [[ 7, 8, 9],
                    [ 10, 11, 12]],
                   [[ 13, 14, 15],
                    [ 16, 17, 18]],
                   [[ 19, 20, 21],
                    [ 22, 23, 24]]])
transposed_tensor = geek.transpose(x, perm = [0, 2, 1])

输出：

array([[[ 1,  4],
        [ 2,  5],
        [ 3,  6]],

       [[ 7, 10],
        [ 8, 11],
        [ 9, 12]],

       [[13, 16],
        [14, 17],
        [15, 18]],

       [[19, 22],
        [20, 23],
        [21, 24]]])
shape (4, 3, 2)

形状是 (4, 3, 2) 因为我们的烫发是 [0, 2, 1]。以下是从 perm 到输入张量形状的映射。

0 => 4
2 => 3
1 => 2

示例 3：现在我们将研究共轭参数
当我们的张量中有复杂的变量时，它被设置为True 。

import tensorflow as geek
  
  
x = geek.constant([[1 + 1j, 2 + 2j, 3 + 3j],
                   [4 + 4j, 5 + 5j, 6 + 6j]])
transposed_tensor = geek.transpose(x)

输出：

array([[1 + 1j, 4 + 4j],
       [2 + 2j, 5 + 5j],
       [3 + 3j, 6 + 6j]])