I have an array:
a = np.array([0,0,0,0,0,0])
I want to add some other array into each index of a, while the index can appear more than one times. I want to get the some of each index. I write:
a[np.array([1,2,2,1,3])] += np.array([1,1,1,1,1])
but get a to be:
array([0, 1, 1, 1, 0, 0])
But what I want is to get:
array([0, 2, 2, 1, 0, 0])
How to implement this in numpy without for loop?
Using pure numpy, AND avoiding a for loop:
np.add.at(a, np.array([1,2,2,1,3]), np.array([1,1,1,1,1]))
Output:
>>> a = np.array([0,0,0,0,0,0])
>>> np.add.at(a, np.array([1,2,2,1,3]), np.array([1,1,1,1,1]))
>>> a
array([0, 2, 2, 1, 0, 0])
Please note, this does in-place substitution. This is what is desired by you, but it may not be desired by future viewers. Hence the note :)
np.add.reduceat but I haven't spent too much time with the ufunc methods so I don't know most of them. The documentation here is even perfect ... "For addition ufunc, this method is equivalent to a[indices] += b, except that results are accumulated for elements that are indexed more than once"
I don't know of a clever numpy vectorized way to do this... the best I can come up with is:
>>> indices = np.array([1,2,2,1,3])
>>> values = np.array([1,1,1,1,1])
>>> a = np.array([0,0,0,0,0,0])
>>> for i, ix in enumerate(indices):
... a[ix] += values[i]
...
>>> a
array([0, 2, 2, 1, 0, 0])
You can do something like (assuming for each index there is a correlated value):
a = np.array([0,0,0,0,0,0])
idxs = np.array([1,2,2,1,3])
vals = np.array([1,1,1,1,1])
for idx, val in zip(idxs,vals):
a[idx] += val
