Sorting an 3-dimensional array by a single row/column in Python - using vectorisation

Question

I have a 3D multidimensional arr = (x, y, z) shaped numpy array. Shape = (10000, 99, 2) in this example.

I.e. we have 10000 instances of 99 x 2 two dimensional arrays.

I would like to sort the whole array by the values in the z index i.e. ranking according to the 99 variables across rows in each column, across each instance.

Is there an easy way to do this using vectorisation? I'm aware I could loop over 10000 iterations, sorting the 2d array like below and combining to a 3d output.

np.unique(arr[:,0], return_inverse=True)
np.unique(arr[:,1], return_inverse=True)

Given I have 10000 outer instances, I am however interested in avoiding loops and sorting all 10000 values in a more efficient manner.

tianlinhe · Accepted Answer · 2020-04-16 09:07:49Z

1

I am not sure if I understand the z score completely, but you can try:

np.sort(arr,axis=1)

An example 3-d input:

import numpy as np
rng_seed = 42  # control reproducibility
rng = np.random.RandomState(rng_seed)
arr=rng.randint(0,40,20).reshape(2,5,2)

The input looks like:

[[[38 28]
  [14  7]
  [20 38]
  [18 22]
  [10 10]]

 [[23 35]
  [39 23]
  [ 2 21]
  [ 1 23]
  [29 37]]]

Applying:

arr1=np.sort(arr,axis=1)
print (arr1)

Gives you the sorted array based on column within each instance:

[[[10  7]
  [14 10]
  [18 22]
  [20 28]
  [38 38]]

 [[ 1 21]
  [ 2 23]
  [23 23]
  [29 35]
  [39 37]]]

If you want the rank of each value instead, try:

arr_rank = arr.argsort(axis=1)
print (arr_rank)

The output is:

[[[4 1]
  [1 4]
  [3 3]
  [2 0]
  [0 2]]

 [[3 2]
  [2 1]
  [0 3]
  [4 0]
  [1 4]]]

edited Apr 16, 2020 at 9:07

answered Apr 16, 2020 at 8:26

tianlinhe

9891 gold badge6 silver badges15 bronze badges

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

2 Comments

StackColumn Over a year ago

thank you, this does indeed sort the values as intended. This may be worth a separate post but is there a similar method for finding the ranked order of each value in its original position in the 3d across the column in each instance.

tianlinhe Over a year ago

Do you want to know the rank of each value? Then, in your case, the output will contains integers 0-98 only? If that is the case check my updated answer.

Collectives™ on Stack Overflow

Sorting an 3-dimensional array by a single row/column in Python - using vectorisation

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related