RuntimeWarning: invalid value encountered in long_scalars

Question

I'm having an issue with modifying an array, by adding the percentage of each item compared to its row to a new matrix. This is the code providing error:

for j in range(1,27):
        for k in range(1,27):
                let_prob[j,k] = let_mat[j,k]*100/(let_mat[j].sum())

I get the error:

RuntimeWarning: invalid value encountered in long_scalars

I have tried rounding the denominator to no success.

Solution

It seems that you are dealing with big numbers, since it raised the error RuntimeWarning. To get rid of such errors, as a numpythonic way you can first calculate the sum of each row using the np.sum() function by specifying the proper axis then repeat and reshape the array in order to be able to divide with your array, them multiple with 100 and round the result:

col, row = np.shape(let_mat) 
let_prob = np.round((let_mat/np.repeat(let_mat.sum(axis=1),row).reshape(col, row).astype(‌​float))*100,2)

Demo :

>>> a = np.arange(20).reshape(4,5)
>>> 
>>> a
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19]])

>>> np.round((a/np.repeat(a.sum(axis=1),5).reshape(4,5).astype(float))*100,2)
array([[  0.  ,  10.  ,  20.  ,  30.  ,  40.  ],
       [ 14.29,  17.14,  20.  ,  22.86,  25.71],
       [ 16.67,  18.33,  20.  ,  21.67,  23.33],
       [ 17.65,  18.82,  20.  ,  21.18,  22.35]])