tf.reshape()与tf.transpose的理解
转载自：https://blog.csdn.net/qq_21735341/article/details/80869615
背景：初次接触tf.transpose，对其中的维度的理解，甚是困难，作此记录，以便以后查看
（1）tf.reshape()的理解
import tensorflow as tf
import numpy as np
three_dim_data = tf.Variable(np.arange(100).reshape(2,5,10))
three_dim_data_reshape = tf.Variable(tf.reshape(three_dim_data,[10,10]))
with tf.Session().as_default() as sess:
sess.run(tf.global_variables_initializer())
print(sess.run(three_dim_data))
print(sess.run(three_dim_data_reshape))
three_dim_data输出结果为：
[[ [ 0 1 2 3 4 5 6 7 8 9]
[10 11 12 13 14 15 16 17 18 19]
[20 21 22 23 24 25 26 27 28 29]
[30 31 32 33 34 35 36 37 38 39]
[40 41 42 43 44 45 46 47 48 49 ]]
[[50 51 52 53 54 55 56 57 58 59]
[60 61 62 63 64 65 66 67 68 69]
[70 71 72 73 74 75 76 77 78 79]
[80 81 82 83 84 85 86 87 88 89]
[90 91 92 93 94 95 96 97 98 99 ]]]
---------------------
three_dim_data_reshape的输出结果为：
[[ 0 1 2 3 4 5 6 7 8 9]
[10 11 12 13 14 15 16 17 18 19]
[20 21 22 23 24 25 26 27 28 29]
[30 31 32 33 34 35 36 37 38 39]
[40 41 42 43 44 45 46 47 48 49]
[50 51 52 53 54 55 56 57 58 59]
[60 61 62 63 64 65 66 67 68 69]
[70 71 72 73 74 75 76 77 78 79]
[80 81 82 83 84 85 86 87 88 89]
[90 91 92 93 94 95 96 97 98 99 ]]
---------------------
通过两种情况的对比，reshape的操作，是将原始数据，先平铺出来[0-99],然后再按照维度的倒序，进行构建数据。例如three_dim_data这是按照，先10，5,2这样的顺序构造数据。three_dim_data_reshape则是先平铺，再10,10这样的顺序构造数据。（
2）tf.transpose的理解
理解了tf.reshape就很容易理解tf.transpose了tf.transpose是改变数据的组成结构。功能与tf.reshape类似。
import tensorflow as tf
import numpy as np
three_dim_data = tf.Variable(np.arange(100).reshape(2,5,10))
three_dim_data_transpose = tf.transpose(three_dim_data,[1,0,2])
transpose_shape = three_dim_data_transpose.shape
with tf.Session().as_default() as sess:
sess.run(tf.global_variables_initializer())
print(sess.run(three_dim_data))
print(sess.run(three_dim_data_transpose))
print(transpose_shape)
输出结果为：[[[ 0 1 2 3 4 5 6 7 8 9]
[50 51 52 53 54 55 56 57 58 59]]
[[10 11 12 13 14 15 16 17 18 19]
[60 61 62 63 64 65 66 67 68 69]]
[[20 21 22 23 24 25 26 27 28 29]
[70 71 72 73 74 75 76 77 78 79]]
[[30 31 32 33 34 35 36 37 38 39]
[80 81 82 83 84 85 86 87 88 89]]
[[40 41 42 43 44 45 46 47 48 49]
[90 91 92 93 94 95 96 97 98 99]]]
(5, 2, 10)
相当于将（2,5,10）reshape为了（5,2,10）tf.transpose()中的[1,0,2]只是在交换（2,5,10）维度的的位置而已，交换后，就可以看成reshape了
