Transpose result of sum

Question

I have two matrices. The calculation of the sum over axis=1. Then I want to transpose the obtained result. The result incorporates in expression. You must be elements of the matrix 1 and 2 calculates the sum, row by row.

Do you need to transpose the array of sum or is there some other way?

matrix1 = [[ 5.  4.  3.  5.  3.]
           [ 5.  7.  8.  2.  2.]
           [ 8.  2.  4.  0.  3.]
           [ 7.  2.  5.  3.  5.]
           [ 3.  1.  3.  0.  2.]]

matrix2 = [[  5.   7.   6.   5.   4.]
           [  3.  45.   2.   3.   4.]
           [  2.   4.   6.   4.   3.]
           [  3.   4.   5.   6.  54.]
           [  4.   3.   6.   7.   5.]]

s_1 = np.array(matrix_1.sum(axis=1))
s_1 = np.array(matrix_2.sum(axis=1))

s_1T = np.transpose(s_1)
s_2T = np.transpose(s_2)

Result:

S_1T
[ 20.  24.  17.  22.   9.]

S_2T
[ 27.  57.  19.  72.  25.]

How to transpose the array?

FROM
[ 20.  24.  17.  22.   9.]
[ 27.  57.  19.  72.  25.]

TO
[20.
 24.
 17.
 22.
  9.]

 [27.
  57.
  19.
  72.
  25.]

Transposed array using in the expression:

result = ((matrix_2 - matrix_1)/matrix_1)/((s_2T-s_1T)/s_1T)

you probably mean to [[20.] [24.] [17.] [22.] [9.]]? – iced Mar 13 '15 at 11:01 — iced, Mar 13 '15 at 11:01

score 1 · Accepted Answer · answered Mar 13 '15 at 11:03

1

s1_t = map(lambda e: [e], s_1)

Most likely it's answer you seek.

answered Mar 13 '15 at 11:03

iced

1,562
8
10

2

This will give you a Python list, not a NumPy array – YXD Mar 13 '15 at 11:14
ye, ok. s1_t = np.array(map(lambda e: [e], s_1)) – kamfulebu Mar 13 '15 at 11:23

Alex Riley · Answer 2 · 2015-03-13T11:22:17.330

1

You're losing an axis when you take the sum and getting back a one-dimensional array. Transposing has no effect on one-dimensional arrays.

You can either insert a new axis, or you can specify keepdims=True when you take the sum to maintain two dimensions (and avoid the need to transpose the array):

matrix1.sum(axis=1, keepdims=True)

returns:

np.array([[20.],
          [24.],
          [17.],
          [22.],
          [ 9.]])

edited Mar 13 '15 at 11:22

answered Mar 13 '15 at 11:06

Alex Riley

169,130
45
262
238

TypeError: 'keepdims' is an invalid keyword argument for this function – kamfulebu Mar 13 '15 at 11:21
Which version of NumPy are you using? – Alex Riley Mar 13 '15 at 11:25
1.6.1 Packaged in ArcGIS 10.2. – kamfulebu Mar 13 '15 at 11:31
2

Ah - you'll need a more recent version of NumPy to use the `keepdims` keyword. In the meantime, `s_1[:,None]` is a good way to put the missing axis back in. – Alex Riley Mar 13 '15 at 11:33

Transpose result of sum

2 Answers2