11/4/24, 3:54 PM numpy - Python array/matrix dimension - Stack Overflow
Python array/matrix dimension
Asked 9 years, 6 months ago Modified 1 year, 9 months ago Viewed 412 times
I create two matrices
1 import numpy as np
arrA = np.zeros((9000,3))
arrB = np.zerros((9000,6))
I want to concatenate pieces of those matrices. But when I try to do:
arrC = np.hstack((arrA, arrB[:,1]))
I get an error:
ValueError: all the input arrays must have same number of dimensions
I guess it's because np.shape(arrB[:,1]) is equal (9000,) instead of (9000,1) , but I cannot
figure out how to resolve it.
Could you please comment on this issue?
python numpy matrix
Share Edit Follow Flag asked Apr 21, 2015 at 17:38
Leopoldo
805 2 11 23
1 I don't really see your issue here. You have a 9000x3 and a 9000x6 array, they are different sizes, so you
cant concat them together... am I missing something? – Mo H. Apr 21, 2015 at 17:39
I want to take 9000x3 array and concatenate it with 9000x1 array getting 9000x4 array. I hope this
comment describe my issue completely. – Leopoldo Apr 21, 2015 at 17:42
Yes the dimensions/size format of the matrices is wrong they need to be equal. Try np.hstack((arrA,
np.zeros((arrayB[1], 3)))) – user3679917 Apr 21, 2015 at 17:43
@Emyen it does not work. – Leopoldo Apr 21, 2015 at 17:51
https://stackoverflow.com/questions/29779401/python-array-matrix-dimension?rq=3 1/4
11/4/24, 3:54 PM numpy - Python array/matrix dimension - Stack Overflow
4 Answers Sorted by: Highest score (default)
You could preserve dimensions by passing a list of indices, not an index:
3 >>> arrB[:,1].shape
(9000,)
>>> arrB[:,[1]].shape
(9000, 1)
>>> out = np.hstack([arrA, arrB[:,[1]]])
>>> out.shape
(9000, 4)
Share Edit Follow Flag answered Apr 21, 2015 at 18:01
DSM
352k 67 601 500
This is easier to see visually.
Assume:
3
>>> arrA=np.arange(9000*3).reshape(9000,3)
>>> arrA
array([[ 0, 1, 2],
[ 3, 4, 5],
[ 6, 7, 8],
...,
[26991, 26992, 26993],
[26994, 26995, 26996],
[26997, 26998, 26999]])
>>> arrB=np.arange(9000*6).reshape(9000,6)
>>> arrB
array([[ 0, 1, 2, 3, 4, 5],
[ 6, 7, 8, 9, 10, 11],
[ 12, 13, 14, 15, 16, 17],
...,
[53982, 53983, 53984, 53985, 53986, 53987],
[53988, 53989, 53990, 53991, 53992, 53993],
[53994, 53995, 53996, 53997, 53998, 53999]])
If you take a slice of arrB, you are producing a series that looks more like a row:
>>> arrB[:,1]
array([ 1, 7, 13, ..., 53983, 53989, 53995])
What you need is a column the same shape as a column to add to arrA:
>>> arrB[:,[1]]
array([[ 1],
[ 7],
https://stackoverflow.com/questions/29779401/python-array-matrix-dimension?rq=3 2/4
11/4/24, 3:54 PM numpy - Python array/matrix dimension - Stack Overflow
[ 13],
...,
[53983],
[53989],
[53995]])
Then hstack works as expected:
>>> arrC=np.hstack((arrA, arrB[:,[1]]))
>>> arrC
array([[ 0, 1, 2, 1],
[ 3, 4, 5, 7],
[ 6, 7, 8, 13],
...,
[26991, 26992, 26993, 53983],
[26994, 26995, 26996, 53989],
[26997, 26998, 26999, 53995]])
An alternate form is to specify -1 in one dimension and the number of rows or cols desired as the
other in .reshape() :
>>> arrB[:,1].reshape(-1,1) # one col
array([[ 1],
[ 7],
[ 13],
...,
[53983],
[53989],
[53995]])
>>> arrB[:,1].reshape(-1,6) # 6 cols
array([[ 1, 7, 13, 19, 25, 31],
[ 37, 43, 49, 55, 61, 67],
[ 73, 79, 85, 91, 97, 103],
...,
[53893, 53899, 53905, 53911, 53917, 53923],
[53929, 53935, 53941, 53947, 53953, 53959],
[53965, 53971, 53977, 53983, 53989, 53995]])
>>> arrB[:,1].reshape(2,-1) # 2 rows
array([[ 1, 7, 13, ..., 26983, 26989, 26995],
[27001, 27007, 27013, ..., 53983, 53989, 53995]])
There is more on array shaping and stacking here
Share Edit Follow Flag edited Jan 15, 2023 at 21:12 answered Apr 21, 2015 at 18:32
Glorfindel dawg
22.6k 13 88 116 103k 23 135 214
I would try something like this:
2 np.vstack((arrA.transpose(), arrB[:,1])).transpose()
https://stackoverflow.com/questions/29779401/python-array-matrix-dimension?rq=3 3/4
11/4/24, 3:54 PM numpy - Python array/matrix dimension - Stack Overflow
Share Edit Follow Flag answered Apr 21, 2015 at 17:56
crispamares
530 4 9
It works and it's a little bit complicated. Thank you for your comment. – Leopoldo Apr 21, 2015 at
19:43
There several ways of making your selection from arrB a (9000,1) array:
2 np.hstack((arrA,arrB[:,[1]]))
np.hstack((arrA,arrB[:,1][:,None]))
np.hstack((arrA,arrB[:,1].reshape(9000,1)))
np.hstack((arrA,arrB[:,1].reshape(-1,1)))
One uses the concept of indexing with an array or list, the next adds a new axis (e.g.
np.newaxis ), the third uses reshape . These are all basic numpy array manipulation tasks.
Share Edit Follow Flag edited Apr 21, 2015 at 18:48 answered Apr 21, 2015 at 18:27
hpaulj
230k 14 253 376
https://stackoverflow.com/questions/29779401/python-array-matrix-dimension?rq=3 4/4
