How can I obtain the same 'special' solutions to underdetermined linear systems that Matlab's `A \ b` (mldivide) operator returns using numpy/scipy?

Question

I found a link where it is shown with an example that the Matlab mldivide operator (\\) gives \'special\' solutions when the system of linear equat

Answer 1

np.array([[8],[18]]).shape

is

(2,1) 

but you want

(2,)

#!/usr/bin/env python3
import numpy as np
from scipy.optimize import nnls
test_A = np.array([[1,2,0],[0,4,3]])

try:
    test_b = np.array([[8],[18]]) # wrong
    print(nnls(test_A,test_b))
except Exception as e:
    print(str(e))

test_b = np.array([8,18]) # sic!
print(nnls(test_A,test_b))

output:

expected vector
(array([ 0.        ,  4.        ,  0.66666667]), 0.0)