How to quickly map multiple numpy array with a function to single array of different object

Question

Say I have 4 numpy arrays like:

a : np.random.randn(4,4)

which is like

array([[ 0.8087178 , -1.32970726, -0.62254544,  0.82266235],
   [-0.57101685, -1.39512136, -0.19650182,  0.46574661],
   [-1.06096896,  0.92744505,  1.98807088,  1.11976449],
   [-0.0103123 ,  1.49460442, -0.16151632,  0.96951708]])

then I have b, c, d also equal np.random.randn(4,4)

Now, suppose I want to know the maximum of for each array location in string, like this:

 array([[a, b, d, c],
        [b, c, a, a],
        [c, a, b, b],
        [d, c, a, b]])

How would I do it? lambda function don't seem to work here. I know a loop would work, but is there a batch function that could do this?

More generally, the question would be, how could I apply individual functions to each array element across arrays without using a loop?

Thanks!

What is 'lambda function'?

hpaulj
– hpaulj

2018-08-07 01:17:47 +00:00
Commented Aug 7, 2018 at 1:17 — hpaulj
– hpaulj, Commented Aug 7, 2018 at 1:17

Julien · Accepted Answer · 2018-08-07 00:34:53Z

2

As simple as this:

np.argmax((a,b,c,d),axis=0)

This will give you an array where 0 means a, 1 means b, and so on. If you really care, you can then just apply the map dict(zip(range(4),'abcd')) to it.

For the "more general" question, you are looking at numpy.vectorize.

answered Aug 7, 2018 at 0:34

Julien

15.3k6 gold badges33 silver badges58 bronze badges

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

1 Comment

sacuL Over a year ago

np.array(['a','b','c','d'])[np.argmax((a,b,c,d),axis=0)] will return the matrix OP was looking for. +1

Oscar Benjamin · Accepted Answer · 2018-08-07 00:38:03Z

For your main question: if you stack the arrays together using np.concatenate then you can use np.argmax to get an array that shows which array has the greatest element in each position. The example below shows this for 1D arrays. You'll want axis=2 for 2D arrays.

In [1]: import numpy as np

In [2]: a1 = np.array([1, 2, 3, 4])

In [3]: a2 = np.array([3, 2, 1, 1])

In [4]: a1
Out[4]: array([1, 2, 3, 4])

In [5]: a2
Out[5]: array([3, 2, 1, 1])

In [6]: np.concatenate([a1, a2])
Out[6]: array([1, 2, 3, 4, 3, 2, 1, 1])

In [7]: np.concatenate([[a1], [a2]])
Out[7]: 
array([[1, 2, 3, 4],
       [3, 2, 1, 1]])

In [9]: ai = np.concatenate([[a1], [a2]])

In [11]: np.argmax(ai, axis=1)
Out[11]: array([3, 0])

In [12]: np.argmax(ai, axis=0)
Out[12]: array([1, 0, 0, 0])

Your other question "how could I apply individual functions to each array element across arrays without using a loop?" is unclear. I don't understand what it means but it should probably be a separate question.

Collectives™ on Stack Overflow

How to quickly map multiple numpy array with a function to single array of different object

2 Answers 2

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related