Numpy, apply a list of functions along array dimension

Question

I have a list of functions of the type:

func_list = [lambda x: function1(input),
             lambda x: function2(input),
             lambda x: function3(input),
             lambda x: x]

and an array of shape [4, 200, 200, 1] (a batch of images).

I want to apply the list of functions, in order, along the 0th axis.

EDIT: Rephrasing the problem. This is equivalent to the above. Say, instead of the array, I have a tuple of 4 identical arrays, of shape (200, 200, 1), and I want to apply function1 on the first element, function2 on the second element, etc. Can this be done without a for loop?

Just to understand, you want to apply it over a particular row? Axis? Or over the entire matrix? Do these functions transform each element or are they something like sum/max functions which give a single output? — cs95
– cs95, Commented Jun 18, 2017 at 11:30

cs95 · Accepted Answer · 2017-06-18 11:33:19Z

3

You can iterate over your function list using np.apply_along_axis:

import numpy as np
x = np.ranom.randn(100, 100)
for f in fun_list:
    x = np.apply_along_axis(f, 0, x)

Based on OP's Update

Assuming your functions and batches are the same in size:

batch = ... # tuple of 4 images  
batch_out = tuple([np.apply_along_axis(f, 0, x) for f, x in zip(fun_list, batch)])

edited Jun 18, 2017 at 11:33

answered Jun 18, 2017 at 11:00

cs95

406k106 gold badges744 silver badges797 bronze badges

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

2 Comments

Qubix Over a year ago

Is it possible to avoid a for loop ?

cs95 Over a year ago

@Qubix It'd be a good feature to have if there was a way. But I'm not sure there is.

Qubix · Accepted Answer · 2017-06-18 12:14:39Z

2

I tried @Coldspeed's answer, and it does work (so I will accept it) but here is an alternative I found, without using for loops:

result = tuple(map(lambda x,y:x(y), functions, image_tuple))

Edit: forgot to add the tuple(), thanks @Coldspeed

edited Jun 18, 2017 at 12:14

answered Jun 18, 2017 at 11:35

Qubix

4,3737 gold badges43 silver badges88 bronze badges

2 Comments

cs95 Over a year ago

If using python 3, don't forget to tack on tuple() to the final answer.

hpaulj Over a year ago

[x(y) for x, y in zip(functions, image_tuple)] is the comprehension equivalent.

Collectives™ on Stack Overflow

Numpy, apply a list of functions along array dimension

2 Answers 2

Based on OP's Update

2 Comments

2 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Based on OP's Update

2 Comments

2 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related