How to find index and max value for overlapping numpy arrays

Question

I have two numpy Array with the same SHAPEs. One with values and one with "zones". I need to find max value and index of the value in valuearr which overlap zone 3 in zonearr:

import numpy as np
valuearr = np.array([[10,11,12,13],
                     [21,22,23,24],
                     [31,32,33,34],
                     [41,42,43,44]])

zonearr = np.array([ [0,0,1,1],
                     [0,0,1,1],
                     [3,3,0,0],
                     [3,3,0,0]])

Im trying:

valuearr[np.argwhere(zonearr==3)].max()
44

When it should be 42.

To get index i try

ind = np.unravel_index(np.argmax(valuearr[np.argwhere(zonearr==3)], axis=None), valuearr.shape)

Which of course doesnt work since max value is not 44 and also give error:

builtins.ValueError: index 19 is out of bounds for array with size 16

For me print(valuearr[zonearr==3].max()) gives 42. Maybe make sure that your zonearr is of dtype=np.int. — Thomas Kühn
– Thomas Kühn, Commented Feb 8, 2019 at 10:01
Oh, I think I only start to understand now: did you also want to find the correct index of the value 42? — Thomas Kühn
– Thomas Kühn, Commented Feb 8, 2019 at 10:04
Yes that is giving me 42 which is correct. Yes i also want the index of max value — Bera
– Bera, Commented Feb 8, 2019 at 10:04

Thomas Kühn · Accepted Answer · 2019-02-08 10:13:53Z

2

You can use a masked array to do what you want.

With:

import numpy as np
valuearr = np.array([[10,11,12,13],
                     [21,22,23,24],
                     [31,32,33,34],
                     [41,42,43,44]])

zonearr = np.array([ [0,0,1,1],
                     [0,0,1,1],
                     [3,3,0,0],
                     [3,3,0,0]], dtype=np.int)

First mask out all the values where zonearr is not equal to 3:

masked = np.ma.masked_array(valuearr, mask = (zonearr!=3))

Then find the position of the maximum value with argmax:

idx_1d = np.argmax(masked)

Finally, convert it into a 2d index:

idx_2d = np.unravel_index(idx_1d, valuearr.shape)

and print:

print(idx_2d, valuearr[idx_2d])

which gives:

(3, 1) 42

answered Feb 8, 2019 at 10:13

Thomas Kühn

9,9003 gold badges53 silver badges66 bronze badges

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

1 Comment

Ashwin Geet D'Sa Over a year ago

@BERA, please check this solution, it works perfectly.

Ashwin Geet D'Sa · Accepted Answer · 2019-02-08 10:10:37Z

1

Please try the below code

np.max(valuearr[np.where(zonearr==3)])

It fetches the indices of the elements from zonearr, where the value equals to '3'. Followed by, obtaining the maximum element from valuearr through the obtained indices.

To obtain the index of the element 42(as per your example), please use the below code:

np.argwhere(valuearr==np.max(valuearr[np.where(zonearr==3)]))

edited Feb 8, 2019 at 10:10

answered Feb 8, 2019 at 10:06

Ashwin Geet D'Sa

7,4583 gold badges40 silver badges66 bronze badges

4 Comments

Ashwin Geet D'Sa Over a year ago

I don't see any differences in the results. But, just the differences in the approach used.

Ashwin Geet D'Sa Over a year ago

Simpler way: maxval=valuearr[zonearr==3].max() followed by index=index=np.argwhere(valuearr==maxval)

Bera Over a year ago

I Think that might find index of larger max values in other zones than 3. (my example data is not representing the real data)

Ashwin Geet D'Sa Over a year ago

@BERA, please look at the below solution, i.e, using a masked array, which works perfectly.

Collectives™ on Stack Overflow

How to find index and max value for overlapping numpy arrays

2 Answers 2

1 Comment

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related