numpy array reshape adding dimension

Question

Ok I am a (very) novice Python user, but I am trying to translate a piece of Python code into R, and I have run into a confusing problem with array reshaping.

Lets make some example data:

X1 = np.array([[-0.047, -0.113, 0.155, 0.001],
        [0.039, 0.254, 0.054, 0.201]], dtype=float)

In:X1
Out:
array([[-0.047, -0.113, 0.155, 0.001],
        [0.039, 0.254, 0.054, 0.201]])
In:X1.shape
Out: (2,4)

Ok so I've made a 2D array with 2 rows and 4 columns. I'm happy with this. The confusion arises with this line of code:

X2 = X1.reshape((2, -1, 1))

In: X2
Out:
array([[[-0.047],
        [-0.113],
         [0.155],
         [0.001]],

         [0.039],
         [0.254],
         [0.054],
         [0.201]]])

In: X2.shape
Out: (2, 4, 1)

So I know that I have added an extra dimension (which I think is the 3rd digit 1 in the reshape command), but I don't understand what else this had done. The shape implies it is still got 2 rows and 4 columns, but clearly something else is changed. Again my motivation here is to do the same operation in R, but until I know I understand what I've transformed here I am stuck. (Forgive me if this is an awful question I only started Python yesterday!)

blue note · Accepted Answer · 2017-01-25 12:29:40Z

5

By reshape(2, -1, 1) you have not just added added a new dimension. You have said

* the 1st dimension should be of size 2
* the 3rd dimension should be of size 1
* the 2nd should be whatever remains

so, the only valid option if 4. If you just want to add a new dimension to an existing matrix, you should do something like x[:, np.newaxis, :] (exact usage depends on what you want the output format to be)

answered Jan 25, 2017 at 12:29

blue note

29.5k10 gold badges83 silver badges110 bronze badges

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

3 Comments

user2498193 Over a year ago

Thats really helpful thank you. In my example does since X1 had dimension (2, 4) and X2 has dimension (2,4,1) - why when I print them does X1 shows the rows with cells side by side, while` X2` shows the individual cells one under another?

blue note Over a year ago

Well, printing 3D data in 2D space is confusing indeed. X2 has 3 opening brackets, so the data you get are ordered along the 3rd dimension, not along rows. For example, try printing X2[0, 0, :] the make it clearer.

user2498193 Over a year ago

ok thanks @blue_note - this has been really helpful. Helped me to figure out that the problem is not in these steps, but with later array algebra being handled differently - many thanks!

Jacques Gaudin · Accepted Answer · 2017-01-25 14:38:28Z

3

There are 3 different ways of adding a dimension to a 2D array.

You should try different combinations to understand the use of reshape. Try the following:

import numpy as np

X1 = np.array([[-0.047, -0.113, 0.155, 0.001],
        [0.039, 0.254, 0.054, 0.201]], dtype=float)

X2 = X1.reshape((1, 2, -1))
print(X2)

>[[[-0.047 -0.113  0.155  0.001]
   [ 0.039  0.254  0.054  0.201]]]

X3 = X1.reshape((-1, 1, 2))
print(X3)

>[[[-0.047 -0.113]]
  [[ 0.155  0.001]]
  [[ 0.039  0.254]]
  [[ 0.054  0.201]]]

edited Jan 25, 2017 at 14:38

answered Jan 25, 2017 at 12:33

Jacques Gaudin

17.1k11 gold badges59 silver badges80 bronze badges

Collectives™ on Stack Overflow

numpy array reshape adding dimension

2 Answers 2

3 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

3 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related