What is the difference between concatenate and stack in numpy-CodePudding

I am bit confused between both the methods : concatenate and stack

The concatenate and stack provides exactly same output , what is the difference between both of them ?

Using : concatenate

import numpy as np
my_arr_1 = np.array([ [1,4] ,   [2,7] ])
my_arr_2 = np.array([ [0,5]   , [3,8] ])

join_array=np.concatenate((my_arr_1,my_arr_2),axis=0)
print(join_array)

Using : stack

import numpy as np
my_arr_1 = np.array([ [1,4] ,   [2,7] ])
my_arr_2 = np.array([ [0,5]   , [3,8] ])

join1_array=np.stack((my_arr_1,my_arr_2),axis=0)
print(join1_array)

Output for both is same :

[[[1 4]
  [2 7]]

 [[0 5]
  [3 8]]]

CodePudding user response：

In [160]: my_arr_1 = np.array([ [1,4] ,   [2,7] ])
     ...: my_arr_2 = np.array([ [0,5]   , [3,8] ])
     ...: 
     ...: join_array=np.concatenate((my_arr_1,my_arr_2),axis=0)
In [161]: join_array
Out[161]: 
array([[1, 4],
       [2, 7],
       [0, 5],
       [3, 8]])
In [162]: _.shape
Out[162]: (4, 2)

concatenate joined the 2 arrays on an existing axis, so the (2,2) become (4,2).

In [163]: join1_array=np.stack((my_arr_1,my_arr_2),axis=0)
In [164]: join1_array
Out[164]: 
array([[[1, 4],
        [2, 7]],

       [[0, 5],
        [3, 8]]])
In [165]: _.shape
Out[165]: (2, 2, 2)

stack joined them on a new axis. It actually made them both (1,2,2) shape, and then used concatenate.

The respective docs should make this clear.