Numpy accumulating one array in another using index array

Question

My question is about a specific array operation that I want to express using numpy.

I have an array of floats w and an array of indices idx of the same length as w and I want to sum up all w with the same idx value and collect them in an array v. As a loop, this looks like this:

for i, x in enumerate(w):
     v[idx[i]] += x

Is there a way to do this with array operations? My guess was v[idx] += w but that does not work, since idx contains the same index multiple times.

Thanks!

And atomic really is an inappropriate description for what you want to do. — Constantinius
– Constantinius, Commented Mar 20, 2012 at 16:06
@Constantinius, He accepted an answer for all but one of the questions he asked, which had one response which, while good, did not address his actual problem. Perhaps you should work on not drawing attention to the game aspect of this site. — Mike Graham
– Mike Graham, Commented Mar 20, 2012 at 16:12
I like the new description, much better. Atomic is often referred to in context of multi-threaded programming, so the question title was misleading. — Constantinius
– Constantinius, Commented Mar 20, 2012 at 16:12
@MikeGraham I accepted some good answers as response to his comment. — Andreas Mueller
– Andreas Mueller, Commented Mar 20, 2012 at 16:13

Bi Rico · Accepted Answer · 2012-03-20 16:28:38Z

15

numpy.bincount was introduced for this purpose:

tmp = np.bincount(idx, w)
v[:len(tmp)] += tmp

I think as of 1.6 you can also pass a minlength to bincount.

answered Mar 20, 2012 at 16:28

Bi Rico

25.9k3 gold badges57 silver badges75 bronze badges

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

2 Comments

Andreas Mueller Over a year ago

I was aware of bincount but I didn'know it could handle weights! awesome :)

Andreas Mueller Over a year ago

Please feel free to feel as my hero for the rest of the day :)

Mike Graham · Accepted Answer · 2012-03-20 16:30:54Z

4

This is a known behavior and, though somewhat unfortunate, does not have a numpy-level workaround. (bincount can be used for this if you twist its arm.) Doing the loop yourself is really your best bet.

Note that your code might have been a bit more clear without re-using the name w and without introducing another set of indices, like

for i, w_thing in zip(idx, w):
    v[i] += w_thing

If you need to speed up this loop, you might have to drop down to C. Cython makes this relatively easy.

edited Mar 20, 2012 at 16:30

answered Mar 20, 2012 at 16:21

Mike Graham

77.2k16 gold badges105 silver badges131 bronze badges

4 Comments

Fred Foo Over a year ago

Simpler still: for i in idx: v[i] += w[i].

Andreas Mueller Over a year ago

Without indices this should work, right? The code you posted simply does v + w, right? (If v is longer than w only the first len(w) elements are used. Reusing w was an pretty bad typo, sorry.

Mike Graham Over a year ago

@AndreasMueller, Sorry, I had a typo.

Mike Graham Over a year ago

@larsmans, There was a typo in my code snippet which I have fixed. That code does not do the same thing as OP's code.

Collectives™ on Stack Overflow

Numpy accumulating one array in another using index array

2 Answers 2

2 Comments

4 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

2 Comments

4 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related