Return values from array based on indices of common values in two other arrays

Question

import numpy as np

a=np.random.randint(0,200,100)#rand int array
b1=np.random.randint(0,100,50)
b2=b1**3
c=[]

I have a problem I think should be easy but can't find solution, I want to find the matching values in two arrays, then use the indices of one of these to find values in another array

for i in range(len(a)):
    for j in range(len(b1)):
         if b1[j]==a[i]:
             c.append(b2[j])

c=np.asarray(c)

Clearly the above method does work, but it's very slow, and this is just an example, in the work I'm actually do a,b1,b2 are all over 10,000 elements.

Any faster solutions?

Any faster solutions? codereview is better suited for this — Bhargav Rao
– Bhargav Rao, Commented Apr 14, 2015 at 15:33
@BhargavRao I thinks also stack could be used for short codes ;) — Kasravnd
– Kasravnd, Commented Apr 14, 2015 at 15:40

Alex Riley · Accepted Answer · 2015-04-15 14:53:31Z

6

np.in1d(b1, a) returns a boolean array indicating whether each element of b1 is found in a.

If you wanted to get the values in b2 which corresponded to the indices of common values in a and b1, you could use the boolean array to index b2:

b2[np.in1d(b1, a)]

Using this function should be a lot faster as the for loops are pushed down to the level of NumPy's internal routines.

edited Apr 15, 2015 at 14:53

answered Apr 14, 2015 at 15:41

Alex Riley

178k46 gold badges274 silver badges247 bronze badges

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

Comments

Kasravnd · Accepted Answer · 2015-04-14 16:12:55Z

1

You can use numpy.intersect1d to get the intersection between 1d arrays.Note that when you can find the intersection then you don't need the indices or use them to find themselves again!!!

>>> a=np.random.randint(0,200,100)
>>> b1=np.random.randint(0,100,50)
>>> 
>>> np.intersect1d(b1,a)
array([ 3,  9, 17, 19, 22, 23, 37, 53, 55, 58, 67, 85, 93, 94])

You may note that using intersection is a more efficient way as for a[np.in1d(a, b1)] in addition of calling in1d function python is forced to do an extra indexing,for better understanding see the following benchmark :

import numpy as np
s1="""
import numpy as np
a=np.random.randint(0,200,100)
b1=np.random.randint(0,100,50)
np.intersect1d(b1,a)
"""
s2="""
import numpy as np
a=np.random.randint(0,200,100)
b1=np.random.randint(0,100,50)
a[np.in1d(a, b1)]
    """


print ' first: ' ,timeit(stmt=s1, number=100000)
print 'second : ',timeit(stmt=s2, number=100000)

Result:

 first:  3.69082999229
second :  7.77609300613

edited Apr 14, 2015 at 16:12

answered Apr 14, 2015 at 15:38

Kasravnd

108k19 gold badges167 silver badges195 bronze badges

9 Comments

dting Over a year ago

This only gives the intersection not the indices.

dting Over a year ago

I still think you are missing the second part of the question: "I want to find the matching values in two arrays, then use the indices of one of these to find values in another array"

Alex Riley Over a year ago

@Kasra: but the OP doesn't just want the intersecting values, he specifically mentions getting the indices of the intersecting values.

dting Over a year ago

I can think of a few. I'm guessing there is a typo in the original question. c.append(b2[j]) should be the last line in the for loop. This would pretty much indicate the OP is trying to look up a value in B2 that is mapped by index to B1.

Thomas North Over a year ago

Yes it's b2 values I'm looking for

|

Collectives™ on Stack Overflow

Return values from array based on indices of common values in two other arrays

2 Answers 2

Comments

9 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

9 Comments

Your Answer

Sign up or log in

Post as a guest

Related