numpy's argsort on rows work not as expected

Asked Oct 05 '16 at 02:33

Active Oct 05 '16 at 02:33

Viewed 2,307 times

I have a simple python code segment that intended to sort each row of a 2D numpy matrix

>>> import numpy as np
>>> X=np.random.random((3,3))
>>> X
array([[ 0.57758986,  0.76771506,  0.94539164],
       [ 0.37619117,  0.85602478,  0.21527702],
       [ 0.41134957,  0.47028597,  0.0672778 ]])
>>> X.argsort(axis=-1)
array([[0, 1, 2],
       [2, 0, 1],
       [2, 0, 1]], dtype=int64)

What I am trying to do is sorting on each row so that I can get the order of each row from min to max, so I am expecting the output of X.argsort to be:

array([[0, 1, 2],
       [1, 2, 0],
       [1, 2, 0]], dtype=int64)

However definitely argsort does not work in this way, do I have any mis-understanding of the argsort of numpy?

Thanks for any suggestions!

asked Oct 05 '16 at 02:33

Roy Li

Why do you expect `[1, 2, 0]` for the second two rows? – mgilson Oct 05 '16 at 02:36
It looks like you want to *rank* the rows, not `argsort` them. They are not the same thing. – Warren Weckesser Oct 05 '16 at 03:15
See, e.g. http://stackoverflow.com/questions/5284646/rank-items-in-an-array-using-python-numpy – Warren Weckesser Oct 05 '16 at 03:16
second row: 0.37x, 0.85x, 0.21x, should their order, from minimum to maximum, 1, 2, 0? – Roy Li Oct 05 '16 at 03:50
If `X = [0.37, 0.85, 0.21]`, `argsort(X)` gives [2, 0, 1], because the sorted array is `[X[2], X[0], X[1]]`. – Warren Weckesser Oct 05 '16 at 04:02
ok, now I understand, it is really my mis-understanding – Roy Li Oct 05 '16 at 04:11

numpy's argsort on rows work not as expected

0 Answers0