FIlter 2d array using 1d boolean array

Question

I have two numpy arrays.

x = [[1,2], [3,4], [5,6]]
y = [True, False, True]

I'd like to get the element of X of which corresponding element of y is True:

filtered_x = filter(x,y)
print(filtered_x) # [[1,2], [5,6]] should be shown.

I've tried np.extract, but it seems work only when x is 1d array. How do I extract the elements of x corresponding value of y is True?

You can try using a list comprehension like `[val for val in x if y[x.index(val)]]`. Simple and elegant. — Asad Moosvi, Sep 01 '17 at 13:22
@AsadMoosvi and slower than numpy built in functions, and also does not return a np.array... — P. Camilleri, Sep 01 '17 at 14:14

score 11 · Accepted Answer · answered Sep 01 '17 at 13:07

11

Just use boolean indexing:

>>> import numpy as np

>>> x = np.array([[1,2], [3,4], [5,6]])
>>> y = np.array([True, False, True])
>>> x[y]   # or "x[y, :]" because the boolean array is applied to the first dimension (in this case the "rows")
array([[1, 2],
       [5, 6]])

And in case you want to apply it on the columns instead of the rows:

>>> x = np.array([[1,2], [3,4], [5,6]])
>>> y = np.array([True, False])
>>> x[:, y]  # boolean array is applied to the second dimension (in this case the "columns")
array([[1],
       [3],
       [5]])

answered Sep 01 '17 at 13:07

MSeifert

145,886
38
333
352

4

For consistency and clarity, I prefer x[y, :] and x[:, y] – Jblasco Sep 01 '17 at 13:11
2

I tend to prefer to omit trailing `, :` because it's more to type and the result doesn't differ if I include them or omit them. But I can see where it's easier to understand. :) – MSeifert Sep 01 '17 at 13:13

score 0 · Answer 2 · answered Sep 01 '17 at 13:19

0

l=[x[i] for i in range(0,len(y)) if y[i]] this will do it.

answered Sep 01 '17 at 13:19

Vineet Jain

1,515
4
21
31

FIlter 2d array using 1d boolean array

2 Answers2