How to insert extra rows and columns at desired location in a matrix ( python )?

user_18728568 0
Viewed 93

I was running a loop, from where I obtained ( 4 by 4) arrays, and now I want to insert zeros at certain locations in these arrays and turn them into (6 by 6) arrays.

I want to add zeros in third and sixth row , and also 3rd and sixth column

a_obtained = np.array( [[1,     2,   3,    4   ],
                        [5,     6,   7,    8   ],
                        [9,    10,   11,  12   ],
                        [13,   14,   15,   16  ]])

I tried going through np.vstack, np.hstack, but I am not able to place the zeros at specified locations.

a_desired = result = np.array([[1,     2,  0,   3,    4,   0 ],
                               [5,     6,  0,   7,    8,   0 ],
                               [0,     0,  0,   0,    0,   0 ],
                               [9,    10,  0,   11,  12,   0 ],
                               [13,   14,  0,   15,   6,   0 ],
                               [0,     0,  0,   0,    0,   0 ]])
indexing numpy arrays python loops
3 Answers

Using np.insert:

import numpy as np
a_obtained = np.array( [[1,     2,   3,    4   ],
                        [5,     6,   7,    8   ],
                        [9,    10,   11,  12   ],
                        [13,   14,   15,   16  ]])
                        
new_row = np.zeros(4)
a_obtained = np.insert(a_obtained, [3, 4], [new_row], axis= 0)
print(a_obtained)
user_154001630

If we have the desired columns indices, it can be done by np.insert as:

desired_cols_ids = np.array([2, 4], dtype=np.int64)

zero_arr_row = np.zeros((1, a_obtained.shape[1]))
# [[0. 0. 0. 0.]]

a_obtained = np.insert(a_obtained, desired_cols_ids, zero_arr_row, axis=0)
# [[ 1  2  3  4]
#  [ 5  6  7  8]
#  [ 0  0  0  0]
#  [ 9 10 11 12]
#  [13 14 15 16]
#  [ 0  0  0  0]]

zero_arr_col = np.zeros((a_obtained.shape[0], 1))
# [[0.]
#  [0.]
#  [0.]
#  [0.]
#  [0.]
#  [0.]]

a_obtained = np.insert(a_obtained, desired_cols_ids, zero_arr_col, axis=1)
# [[ 1  2  0  3  4  0]
#  [ 5  6  0  7  8  0]
#  [ 0  0  0  0  0  0]
#  [ 9 10  0 11 12  0]
#  [13 14  0 15 16  0]
#  [ 0  0  0  0  0  0]]

For a little explanation, at first, we insert into the array rows (so it is modified) and then insert into the columns of the modified array.

user_133948170

np.insert is relatively easy to use for this task, but internally it uses a variety of methods, depending on the insert specifications. For this case I suspect it does something like the following:

Start with a "blank" array:

In [51]: res = np.zeros((6,6),a_obtained.dtype)

Create a boolean mask of where we want to add rows and columns. Here I can do both rows and columns at once (rather than with 2 insert calls):

In [52]: mask = res.astype(bool)
In [53]: mask[[2,5]] = True
In [54]: mask[:,[2,5]] = True
In [55]: mask
Out[55]: 
array([[False, False,  True, False, False,  True],
       [False, False,  True, False, False,  True],
       [ True,  True,  True,  True,  True,  True],
       [False, False,  True, False, False,  True],
       [False, False,  True, False, False,  True],
       [ True,  True,  True,  True,  True,  True]])

Then just add the original values. Boolean masking selects a "flat" array, so a_obtained has to be flattened as well. Otherwise the insert is straight forward.

In [56]: res[~mask] = a_obtained.ravel()
In [57]: res
Out[57]: 
array([[ 1,  2,  0,  3,  4,  0],
       [ 5,  6,  0,  7,  8,  0],
       [ 0,  0,  0,  0,  0,  0],
       [ 9, 10,  0, 11, 12,  0],
       [13, 14,  0, 15, 16,  0],
       [ 0,  0,  0,  0,  0,  0]])
user_9019250
Related