Replace values in an nd numpy array at given axis index

Question

I want to replace the values in a given numpy array (A) at a given index (e.g. 0), along a given axis (e.g. -2) with a given value (e.g. 0), equivalently:

A[:,:,0,:]=0

Problem is the input array A may come in as 3D or 4D or other shapes, so for 3D data I would need

A[:,0,:]=0

If it's 5D: A[:,:,:,0,:]=0

Currently I'm using an exec() to get this done:

slicestr=[':']*numpy.ndim(var)
slicestr[-2]=str(0)
slicestr=','.join(slicestr)
cmd='A[%s]=0' %slicestr
exec(cmd)
return A

I'm a bit concerned that the usage of exec() might not be quite a nice approach. I know that numpy.take() can give me the column at specific index along specific axis, but to replace values I still need to construct the slicing/indexing string, which is dynamic. So I wonder is there any native numpy way of achieving this?

Thanks.

akuiper · Accepted Answer · 2017-10-16 01:32:43Z

7

You can use Ellipsis (see numpy indexing for more) to skip the first few dimensions:

# assert A.ndim >= 2
A[...,0,:]=0

A2d = np.arange(12).reshape(2,6)
A3d = np.arange(12).reshape(2,3,2)
A4d = np.arange(12).reshape(2,1,3,2)

A2d[...,0,:] = 0

A2d
#array([[ 0,  0,  0,  0,  0,  0],
#       [ 6,  7,  8,  9, 10, 11]])

A3d[...,0,:] = 0

A3d
#array([[[ 0,  0],
#        [ 2,  3],
#        [ 4,  5]],

#       [[ 0,  0],
#        [ 8,  9],
#        [10, 11]]])

A4d[...,0,:] = 0

A4d
#array([[[[ 0,  0],
#         [ 2,  3],
#         [ 4,  5]]],  

#       [[[ 0,  0],
#         [ 8,  9],
#         [10, 11]]]])

edited Oct 16, 2017 at 1:32

answered Oct 16, 2017 at 1:26

akuiper

216k33 gold badges362 silver badges379 bronze badges

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

6 Comments

cs95 Over a year ago

Note to readers: Ellipsis works on python3.x only. For 2.x, you'd need to build a slice object.

Jason Over a year ago

Cool, that's I'm looking for! And it has nothing to do with python3 or 2, it is a numpy thing.

akuiper Over a year ago

@cᴏʟᴅsᴘᴇᴇᴅ Which python 2.x are you using? It seems working for me on python 2.7.13 and numpy 1.12.0.

cs95 Over a year ago

The Ellipsis object exists on 2.x as well, but the ... syntax is invalid, at least on the version I'm using (2.7.10).

Jason Over a year ago

tested and worked in a conda env with python=2.7.10 and numpy=1.12.0, also worked in python=2.6.9 with numpy=1.9.2

|

Collectives™ on Stack Overflow

Replace values in an nd numpy array at given axis index

1 Answer 1

6 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

6 Comments

Your Answer

Sign up or log in

Post as a guest

Related