numpy-axis, expand_dims

• axis=0

 

It will act on all the ROWS in each column, means along "indexes".

It's a row-wise operation.

 

• axis=1

 

It will act on all the COLUMNS in each row, means along "columns".

It's a column-wise operation.

a = np.array([[1, 2]])
b = np.array([[5, 6]])
row_wise=np.concatenate((a, b), axis=0)
column_wise=np.concatenate((a, b), axis=1)

print(row_wise)
print(column_wise)

>>>

[[1 2]
 [5 6]]
 
[[1 2 5 6]]

---

- multi array

m=np.array([[[1,2,3],
            [4,5,6],
            [7,8,9]],
 
            [[10,11,12],
            [13,14,15],
            [16,17,18]]])
 
n=np.array([[[51,52,53],
            [54,55,56],
            [57,58,59]],
 
            [[110,111,112],
            [113,114,115],
            [116,117,118]]])
            
np.array([m,n])
>>>
array([[[[  1,   2,   3],
         [  4,   5,   6],
         [  7,   8,   9]],

        [[ 10,  11,  12],
         [ 13,  14,  15],
         [ 16,  17,  18]]],

       [[[ 51,  52,  53],
         [ 54,  55,  56],
         [ 57,  58,  59]],

        [[110, 111, 112],
         [113, 114, 115],
         [116, 117, 118]]]])

np.array([m,n]).shape
>>>
(2, 2, 3, 3)

 

- axis : integer, the axis along which you want to stack the arrays. -1 means last dimension. e.g. for 2D arrays axis 1 and -1 are same. 

np.array([m,n]).shape
>>>
(2, 2, 3, 3)
axis 0th, 1th, 2nd, 3rd

---

array( [  [  [  [    element  ] ] ] ])

 

case1.

axis=0

np.stack((m,n),axis=0)

 

array([[[[  1,   2,   3],
              [  4,   5,   6],
              [  7,   8,   9]],

          [[ 51,  52,  53],
           [ 54,  55,  56],
           [ 57,  58,  59]]]

 

case2.

axis=1

np.stack((m,n),axis=1)

 

array([[[[  1,   2,   3],
              [  4,   5,   6],
              [  7,   8,   9]],

          [[ 51,  52,  53],
           [ 54,  55,  56],
           [ 57,  58,  59]]]

 

case3.

axis=2

np.stack((m,n),axis=2)

 

array([[[[  1,   2,   3],
              [  4,   5,   6],
              [  7,   8,   9]],

          [[ 51,  52,  53],
           [ 54,  55,  56],
           [ 57,  58,  59]]]

 

case4.

axis=3

np.stack((m,n),axis=3)

 

array([[[[  1,   2,   3],
              [  4,   5,   6],
              [  7,   8,   9]],

          [[ 51,  52,  53],
           [ 54,  55,  56],
           [ 57,  58,  59]]]

 

reference : https://stackoverflow.com/questions/22149584/what-does-axis-in-pandas-mean

---

• expand_dims

 

Expand the shape of an array.

Insert a new axis that will appear at the axis position in the expanded array shape.

x = np.array([1, 2])
y = np.expand_dims(x, axis=(0, 1))
print(y.shape)
print(y)

>>>
(1, 1, 2)
[[[1 2]]]

---

axis=(0, 1) : Put the 1-d in the 0th, 1th
∴
      [0th, 1th, 2th]
(2,)->(1,   1,   2)

[[[1 2]]]

---

axis=(2,0) : Put the 1-d in the 2th, 0th
∴
      [0th, 1th, 2th]
(2,)->(1,    2,   1)

[[[1]
  [2]]]