I have a bidimensional np array V (100000x50). I want to create a new array V_tgt in which I keep just certain rows of V, so the dimension will be (ix50). It may be easy to do it but I tried different things and it seems to save just the first of the 50 elements. My code is the following:
V_tgt = np.array([])
for i in IX_items:
if i in IX_tgt_items:
V_tgt=np.append(V_tgt, V[i])
I tried with functions such as insert and delete as well but it didn't work.How can I save all the values and create an array with the right dimension? Any help is really appreciated.
From your comments I assume that you have some kind of list of target indices (in my example tgt_idx1
and tgt_idx2
)that tells you which elements to take from V. You could do something like this:
import numpy as np
V = np.array([[1,2,3], [4,5,6], [7,8,9], [10, 11, 12]])
tgt_idx1 = np.array([1, 2, 3])
tgt_idx2 = np.array([1, 3])
mask = []
for i, elem in enumerate(V):
inTargets = i in tgt_idx1 and i in tgt_idx2
mask.append(inTargets)
print mask
V_tgt = V[mask]
print V_tgt
This prints
[False, True, False, True]
[[ 4 5 6]
[10 11 12]]
Let's first simplify the problem. With an initial array
( a
):
a = np.array([45, 29, 76, 23, 76, 98, 21, 63])
and the index
arrays
:
i1 = np.array([1, 3, 5, 7, 9])
i2 = np.array([0, 1, 2, 3, 4])
then we can use a simple list
comprehension to get the elements from a
that are at indexes
in both i1
and i2
:
np.array([e for i, e in enumerate(a) if i in i1 and i in i2])
which is very readable and outputs:
array([29, 23])
I am sure you can adapt this to the variables
you have given to your arrays
.
The performance of np.append is probably killing this, why not create a new overlap of your two indices, then subset:
#using @Joe Iddons data
a = np.array([45, 29, 76, 23, 76, 98, 21, 63])
i1 = np.array([1, 3, 5, 7, 9])
i2 = np.array([0, 1, 2, 3, 4])
Then find the intersect of i1 and i2:
indices = np.intersect1d(i1,i2)
and subset:
a[indices]
array([29, 23])
The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.