How to remove a column in a numpy array?

Question

Imagine we have a 5x4 matrix. We need to remove only the first dimension. How can we do it with numpy?

array([[  0.,   1.,   2.,   3.],
       [  4.,   5.,   6.,   7.],
       [  8.,   9.,  10.,  11.],
       [ 12.,  13.,  14.,  15.],
       [ 16.,  17.,  18.,  19.]], dtype=float32)

I tried:

arr = np.arange(20, dtype=np.float32)
matrix = arr.reshape(5, 4)
new_arr = numpy.delete(matrix, matrix[:,0])
trimmed_matrix = new_arr.reshape(5, 3)

It looks a bit clunky. Am I doing it correctly? If yes, is there a cleaner way to remove the dimension without reshaping?

You want to end up with a (5, 3) array? Then you want to delete a column (or in general, an 'entry' from a dimension). Removing a dimension would be changing to a (5,) or a (4,) array. — DilithiumMatrix
– DilithiumMatrix, Commented Nov 30, 2015 at 20:48
It seems you want to remove the first column from a 2D array. This can be done like this: arr[:,1:]. — Christian K.
– Christian K., Commented Nov 30, 2015 at 20:48
np.delete works by index, not value. It is not list remove. — hpaulj
– hpaulj, Commented Nov 30, 2015 at 21:38

Back2Basics · Accepted Answer · 2015-12-01 18:54:00Z

57

If you want to remove a column from a 2D Numpy array you can specify the columns like this

to keep all rows and to get rid of column 0 (or start at column 1 through the end)

a[:,1:]

another way you can specify the columns you want to keep ( and change the order if you wish) This keeps all rows and only uses columns 0,2,3

a[:,[0,2,3]]

The documentation on this can be found here

And if you want something which specifically removes columns you can do something like this:

idxs = list.range(4)
idxs.pop(2) #this removes elements from the list
a[:, idxs]

and @hpaulj brought up numpy.delete()

This would be how to return a view of 'a' with 2 columns removed (0 and 2) along axis=1.

np.delete(a,[0,2],1)

This doesn't actually remove the items from 'a', it's return value is a new numpy array.

edited Dec 1, 2015 at 18:54

answered Nov 30, 2015 at 21:25

Back2Basics

7,8462 gold badges35 silver badges50 bronze badges

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

1 Comment

hpaulj Over a year ago

Only [:,1:] returns a view (i.e. uses the same data buffer). The others make a copy.

hpaulj · Accepted Answer · 2015-11-30 21:41:21Z

9

The correct way to use delete is to specify index and dimension, eg. remove the 1st (0) column (dimension 1):

In [215]: np.delete(np.arange(20).reshape(5,4),0,1)
Out[215]: 
array([[ 1,  2,  3],
       [ 5,  6,  7],
       [ 9, 10, 11],
       [13, 14, 15],
       [17, 18, 19]])

other expressions that work:

np.arange(20).reshape(5,4)[:,1:]
np.arange(20).reshape(5,4)[:,[1,2,3]]
np.arange(20).reshape(5,4)[:,np.array([False,True,True,True])]

answered Nov 30, 2015 at 21:41

hpaulj

233k14 gold badges260 silver badges392 bronze badges

Comments

yevgeniy · Accepted Answer · 2015-11-30 21:40:56Z

4

You don't need the second reshape.

matrix=np.delete(matrix,0,1)

answered Nov 30, 2015 at 21:40

yevgeniy

9288 silver badges14 bronze badges

Collectives™ on Stack Overflow

How to remove a column in a numpy array?

3 Answers 3

1 Comment

Comments

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

1 Comment

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related