Numpy 2D array Subtraction by row

Question

I am struggling to vectorize the following operation. I have an array of x,y,z distances and I need to find the differences between each vector from one another.

temp_result = np.array([[0.8, 0., 1.], [0., -0.6, 1.],[0.8, 0., 1.]])

What I intend to do is subtract without using for loop iteration.

 temp_result[0] - temp_result[0]
 temp_result[0] - temp_result[1]
 temp_result[0] - temp_result[2]
 temp_result[1] - temp_result[0]
 temp_result[1] - temp_result[1]
 temp_result[1] - temp_result[2]
 temp_result[2] - temp_result[0]
 temp_result[2] - temp_result[1]
 temp_result[2] - temp_result[2]

thanks!

nneonneo · Accepted Answer · 2012-10-10 06:31:21Z

3

Here's a nice reshape-based trick:

arr = temp_result
diffs = arr[:,None,:] - arr[None,:,:]

Then the vector difference between arr[i] and arr[j] is found in diffs[i,j].

edited Oct 10, 2012 at 6:31

answered Oct 10, 2012 at 5:49

nneonneo

181k37 gold badges331 silver badges412 bronze badges

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

6 Comments

Warren Weckesser Over a year ago

Here's a more concise version: diffs = arr[:, None] - arr[None, :]. This is an example of numpy's broadcasting capability (docs.scipy.org/doc/numpy/user/basics.broadcasting.html).

nneonneo Over a year ago

I'm exploiting broadcasting already; what you did there was to use newaxis implicitly (which is a nice solution). I'll add it to the answer, thanks!

Warren Weckesser Over a year ago

Sorry, I didn't mean to imply that you weren't using broadcasting! That statement was meant as a general comment about the method.

Fatoons Over a year ago

Thank you both Warren and Nneoneo...could you give any tips on how to no sum the results but not over the whole array, rather only summing the results row by row?

Fatoons Over a year ago

In other words...the above results in an 3X3 array...is there a way I could sum up each entry individually? [0][0] [0][1] [0][2]....

|

Bitwise · Accepted Answer · 2012-10-10 03:56:55Z

0

Check out scipy.spatial.distance, you have functions for all vs all distances.

answered Oct 10, 2012 at 3:56

Bitwise

7,8357 gold badges39 silver badges53 bronze badges

1 Comment

Fatoons Over a year ago

I was looking to stay within the numpy library. Any tips?

Collectives™ on Stack Overflow

Numpy 2D array Subtraction by row

2 Answers 2

6 Comments

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

6 Comments

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related