Numpy - row-wise normalization

Question

I've been working on a matrix normalization problem, stated as:

Given a matrix M, normalize its elements such that each element is divided with the corresponding column sum if element is not 0.

cwsums = np.sum(class_matrix,axis=1)
cwsums = np.reciprocal(cwsums.astype(np.float32))
cwsums[cwsums == np.inf] = 0
## this is the problem
final_matrix = np.multiply(final_matrix, cwsums)

I can construct a reciprocal mask, which I would like to apply accross the matrix, as an elementwise product, yet I can't seem to get it right. Thank you!

EDIT: what I have written seems to work, but it is slow check again, it seems to be row-wise broadcast, not columnwise. — cs95
– cs95, Commented May 14, 2018 at 4:41
So, you want to normalise across columns, correct? You will need axis=1, which is right. Also, my code works for me. What is the shape of your data? — cs95
– cs95, Commented May 14, 2018 at 4:51

cs95 · Accepted Answer · 2018-05-14 05:00:42Z

2

(Addressing edited question) Looks like you meant to sum across rows using axis=0:

i = 1 / class_matrix.sum(axis=0)
i[~np.isfinite(i)] = 0

class_matrix *= i

edited May 14, 2018 at 5:00

answered May 14, 2018 at 4:34

cs95

406k106 gold badges744 silver badges797 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Numpy - row-wise normalization

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related