I am translating code from MATLAB to Python. I need to extract the lower subdiagonal values of a matrix. My attempt in python seems to extract the same values (sum is equal), but in different order. This is a problem as I need to apply corrcoef after.
The original Matlab code is using an array of indices to subset a matrix.
MATLAB code:
values = 1:100;
matrix = reshape(values,[10,10]);
subdiag = find(tril(ones(10),-1));
matrix_subdiag = matrix(subdiag);
subdiag_sum = sum(matrix_subdiag);
disp(matrix_subdiag(1:10))
disp(subdiag_sum)
Output:
2
3
4
5
6
7
8
9
10
13
1530
My attempt in Python
import numpy as np
matrix = np.arange(1,101).reshape(10,10)
matrix_t = matrix.T #to match MATLAB arrangement
matrix_subdiag = matrix_t[np.tril_indices((10), k = -1)]
subdiag_sum = np.sum(matrix_subdiag)
print(matrix_subdiag[0:10], subdiag_sum))
Output: [2 3 13 4 14 24 5 15 25 35] 1530
How do I get the same order output? Where is my error?
Thank you!
CodePudding user response:
For the sum use directly numpy.triu
on the non-transposed matrix:
S = np.triu(matrix, k=1).sum()
# 1530
For the indices, numpy.triu_indices_from
and slicing as a flattened array:
idx = matrix[np.triu_indices_from(matrix, k=1)]
output:
array([ 2, 3, 4, 5, 6, 7, 8, 9, 10, 13, 14, 15, 16, 17, 18, 19, 20,
24, 25, 26, 27, 28, 29, 30, 35, 36, 37, 38, 39, 40, 46, 47, 48, 49,
50, 57, 58, 59, 60, 68, 69, 70, 79, 80, 90])