How to find the index of n largest elements in a list or np.array, Python

Question

Is there a built-in function or a very simple way of finding the index of n largest elements in a list or a numpy array?

K = [1,2,2,4,5,5,6,10]

Find the index of the largest 5 elements?

I count the duplicates more than once, and the output should be a list of the indices of those largest numbers

Answer 1

Maybe something like:

>>> K
[4, 5, 1, 6, 2, 5, 2, 10]
>>> sorted(range(len(K)), key=lambda x: K[x])
[2, 4, 6, 0, 1, 5, 3, 7]
>>> sorted(range(len(K)), key=lambda x: K[x])[-5:]
[0, 1, 5, 3, 7]

or using numpy, you can use argsort:

>>> np.argsort(K)[-5:]
array([0, 1, 5, 3, 7])

argsort is also a method:

>>> K = np.array(K)
>>> K.argsort()[-5:]
array([0, 1, 5, 3, 7])
>>> K[K.argsort()[-5:]]
array([ 4,  5,  5,  6, 10])

Answer 2

Consider the following code,

 N=5
 K = [1,10,2,4,5,5,6,2]
 #store list in tmp to retrieve index
 tmp=list(K)
 #sort list so that largest elements are on the far right
 K.sort()
 #To get the 5 largest elements
 print K[-N:]
 #To get the 5th largest element
 print K[-N]
 #get index of the 5th largest element
 print tmp.index(K[-N])

If you wish to ignore duplicates, then use set() as follows,

 N=5
 K = [1,10,2,4,5,5,6,2]
 #store list in tmp to retrieve index
 tmp=list(K)
 #sort list so that largest elements are on the far right
 K.sort()
 #Putting the list to a set removes duplicates
 K=set(K)
 #change K back to list since set does not support indexing
 K=list(K)
 #To get the 5 largest elements
 print K[-N:]
 #To get the 5th largest element
 print K[-N]
 #get index of the 5th largest element
 print tmp.index(K[-N])

Hopefully one of them covers your question :)

Answer 3

For efficiency, numpy partition is much more efficient for large array. There is no need to sort fully.

"All elements smaller than the k-th element are moved before this element and all equal or greater are moved behind it. The ordering of the elements in the two partitions is undefined."

Answer 4

import headq

Then use function nlargest()