initial a (N+1)-D array from a N-D array with a "value to 1-D array" function

Question

I have a function to convert a value to a list (or array). How can I use it to convert a N-D array to another (N+1)-D array? The code below needs to generate a list then convert to <numpy.ndarray>. I'm wonder if there is a more efficient way (maybe some constructor for <numpy.ndarray>?).

# # assume we have a function foo
# def foo(val):
#     return [val*10, val*20 - 7]
# # how can we use this function to convert a N-D array to another (N+1)-D array?
# # here is my tried 

import numpy as np
def foo(val):
    return [val*10, val*20 - 7]

a = np.arange(3) # 'numpy.ndarray', [0 1 2]
b = [foo(val) for val in a]

print (type(b)) # not good, it is a 'list'
print (b) # [[0, -7], [10, 13], [20, 33]]

b = np.array(b).reshape((3, 2))
print (type(b)) # this is what I want 'numpy.ndarray'
print (b) # will be what I want:
          # [[ 0 -7]
          #  [10 13]
          #  [20 33]]

If there is no <numpy.ndarray> constructor for this. Should I generate a new ndarray (maybe with "numpy.zeros") and fill the value? Or it is recommend to generate a list then convert it to a ndarray? Not sure which one is faster.

Thanks!

Jussi Nurminen · Accepted Answer · 2020-09-25 10:31:54Z

1

You can directly return a numpy array from your function:

def foo(val):
    return np.column_stack((val*10, val*20 - 7))

Because val*10 invokes elementwise multiplication (special case of numpy broadcasting), the result will be another array of the same shape. The same applies to val*20 - 7. Then column_stack() just attaches the two arrays together to get the final output shape you want.

For the general case (N-D to (N+1)-D), you can use:

def foo(val):
    return np.stack((val*10, val*20 - 7), axis=-1)

which will stack your N-D arrays into a (N+1)-D array along the last dimension.

edited Sep 25, 2020 at 10:31

answered Sep 25, 2020 at 9:19

Jussi Nurminen

2,4081 gold badge12 silver badges18 bronze badges

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

2 Comments

Garlic Tseng Over a year ago

Good point! I just tried and I found it works on 1-D array to 2-D array. However if the val is 2-D array the foo function won't return a 3-D array, still studying why...

Jussi Nurminen Over a year ago

@GarlicTseng I edited the answer for the general N-D case

Collectives™ on Stack Overflow

initial a (N+1)-D array from a N-D array with a "value to 1-D array" function

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related