Numpy: A new tremandous BUG in Advanced Indexing?
This is a BUG
train = np.zeros([26,26])
train2 = np.zeros([26,26])
tmp = [1,2,3,4,5,6,1,2,3,4,5,6]
train[tmp[0:-1], tmp[1:]] += 1
for i in range(len(tmp)-1):
train2[tmp[i], tmp[i+1]] += 1
print(np.sum(train - train2))
the output is -5
I find that train is not equal to train2 when there are some duplicate tuples in tmp. Is there something wrong in numpy slicing.
53 - Invalid
Levstyle
👍1
All 3 comments
The same to you.
RogerMonkey
on 30 Mar 2017
You know that you're trying to write to the "same position" twice, for example [1, 2]. Unfortunatly (or fortunatly, depending on the point of view) that doesn't work.
But you can use np.add.at which according to the docs "For addition ufunc, this method is equivalent to a[indices] += b, except that results are accumulated for elements that are indexed more than once.":
train = np.zeros([26,26])
np.add.at(train, [tmp[:-1], tmp[1:]], 1)
MSeifert04
on 30 Mar 2017
👍3
Thanks!
Levstyle
on 30 Mar 2017
Was this page helpful?
0 / 5 - 0 ratings
Related issues
Foadsf
·
3Comments
thouis
·
4Comments
keithbriggs
·
3Comments
qualiaa
·
3Comments
astrofrog
·
4Comments
Most helpful comment
You know that you're trying to write to the "same position" twice, for example
[1, 2]. Unfortunatly (or fortunatly, depending on the point of view) that doesn't work.But you can use
np.add.atwhich according to the docs "For addition ufunc, this method is equivalent to a[indices] += b, except that results are accumulated for elements that are indexed more than once.":