numpy.column_stack vectorized

Question

I would like to vectorize some code, but I find it hard to apply some functions to vectors of variables.

For example, I have two constant vectors, a and b, and a vector of vectors x (a matrix). Dimensions of the members of x is the same as those of a and b. I want to make matrices formed by columns from : a member of x, a, and b:

x = [[ 0.76662363 -0.0397725   0.64086377]
 [ 0.76198581 -0.04605764  0.6459538 ]]

a = [ 0.2763932   0.85065081 -0.5527864 ]

b = [-0.7236068   0.52573111 -0.5527864 ]

The output should be a vector (or array) of two 3x3 matrices. I am trying to run the following code:

a = np.column_stack((x, a, b))

however I get an error message that the dimensions do not match across the arguments:

File "/software/python/2.7.8/lib/python2.7/site-packages/numpy/lib/shape_base.py", line 317, in column_stack
    return _nx.concatenate(arrays, 1)
ValueError: all the input array dimensions except for the concatenation axis must match exactly

Any ideas?

Please add sample input and the expected output. Furthermore add the stacktrace of your error. — ppasler
– ppasler, Commented Feb 11, 2017 at 18:16
How do you rearrange 12 terms (2x3,3,3) into 18 terms (2 3x3)? — hpaulj
– hpaulj, Commented Feb 11, 2017 at 18:46

Paul Panzer · Accepted Answer · 2017-02-11 19:32:38Z

1

Going with the updated requirements:

>>> N = 2    # n cols of x.T
>>> K = 3    # x and a and b
>>> M = 3    # len(a)
>>> outstack = np.empty((N, M, K))
>>> outstack[..., 0] = x
>>> outstack[..., 1] = a[None, :]
>>> outstack[..., 2] = b[None, :]

outstack stacks two 3x3 matrices.

The trick here is preallocating the output stack. The output serves as a broadcasting reference, so a and b can be used almost unchanged and will be broadcast correctly.

edited Feb 11, 2017 at 19:32

answered Feb 11, 2017 at 18:21

Paul Panzer

53.3k3 gold badges60 silver badges103 bronze badges

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

13 Comments

John Over a year ago

Unfortunately, this gives me a single matrix which has dimension 3 x 4, and whose two first columns are x.T, and the latter two columns are a and b. What I want is two 3x3 matrices, one for each element of x.

Paul Panzer Over a year ago

@John you mean you want two copies of x.T one with a joined as the third column, one with b?

John Over a year ago

No. I need two 3x3 matrices. If x1 and x2 are the two rows of x (columns of x.T), the two matrices I need are: (x1 a b) (x2 a b) The reason I need to vectorize the code is that the members of x are 10^8, instead of 2.

hpaulj Over a year ago

John - you should have put that basic information in the question the first time.

John Over a year ago

hpaulj, noted. I assumed that using the term vectorize would do the trick.

|

Collectives™ on Stack Overflow

numpy.column_stack vectorized

1 Answer 1

13 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

13 Comments

Your Answer

Sign up or log in

Post as a guest

Related