>>> x = np.array([[1,2,3],[4,5,6]])
>>> x.shape
(2, 3)
>>> np.expand_dims(x, 0)
array([[[1, 2, 3],
[4, 5, 6]]])
>>> np.expand_dims(x, 1)
array([[[1, 2, 3]],
[[4, 5, 6]]])
>>> np.expand_dims(x, 2)
array([[[1],
[2],
[3]],
[[4],
[5],
[6]]])
>>> np.expand_dims(x, 3)
array([[[1],
[2],
[3]],
[[4],
[5],
[6]]])
>>> np.expand_dims(x, 4)
array([[[1],
[2],
[3]],
[[4],
[5],
[6]]])
>>> np.expand_dims(x, 1).shape
(2, 1, 3)
>>> np.expand_dims(x, 2).shape
(2, 3, 1)
>>> np.expand_dims(x, 3).shape
(2, 3, 1)
>>> np.expand_dims(x, 4).shape
(2, 3, 1)
>>> np.expand_dims(x, 0).shape
(1, 2, 3)
1. 分3部分来看。
2. expand_dims直观来说就是将某一维度展成1维，看shape的形式便可知。
3. 哪一维要设置成1，就将原始DATA重新组合，细心看输出的数据的重排形式。
4. 维度从0开始，超过都表示最后一维，-1也是最后一维。
原文：https://www.cnblogs.com/alexYuin/p/8849526.html