numpy-discussion

[Numpy-discussion] addressing a submatrix

[Martin W. <mar...@gm...>]

Hi list,

is there a concise way to address a subrectangle of a 2d array? So far I'm 
using

A [I] [:, J]

which is not pretty and more importantly only works for reading the 
subrectangle. Writing does *not* work. (Cf. session below.)

Any help would be appreciated.

Thanks,
Martin



In [1]:a = zeros ((4,4))

In [2]:b = ones ((2,2))

In [3]:c = array ((1,2))

In [4]:a [c] [:, c] = b

In [5]:a
Out[5]:
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0]])

In [6]:a [:, c] [c] = b

In [7]:a
Out[7]:
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0]])

In [8]:a [c, c] = b

In [9]:a
Out[9]:
array([[0, 0, 0, 0],
       [0, 1, 0, 0],
       [0, 0, 1, 0],
       [0, 0, 0, 0]])

In [10]:a [c] [:, c]
Out[10]:
array([[1, 0],
       [0, 1]])

In [11]:

Re: [Numpy-discussion] addressing a submatrix

[Simon B. <si...@ar...>]

On Wed, 14 Jun 2006 11:14:17 +0200
Martin Wiechert <mar...@gm...> wrote:

> 
> Hi list,
> 
> is there a concise way to address a subrectangle of a 2d array? So far I'm 
> using
> 
> A [I] [:, J]

what about A[I,J] ?

Simon.

>>> import numpy
>>> a=numpy.zer
numpy.zeros       numpy.zeros_like  
>>> a=numpy.zeros([4,4])
>>> a
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 0, 0]])
>>> a[2:3,2:3]=1
>>> a
array([[0, 0, 0, 0],
       [0, 0, 0, 0],
       [0, 0, 1, 0],
       [0, 0, 0, 0]])
>>> a[1:3,1:3]=1
>>> a
array([[0, 0, 0, 0],
       [0, 1, 1, 0],
       [0, 1, 1, 0],
       [0, 0, 0, 0]])
>>> 


-- 
Simon Burton, B.Sc.
Licensed PO Box 8066
ANU Canberra 2601
Australia
Ph. 61 02 6249 6940
http://arrowtheory.com 

Re: [Numpy-discussion] addressing a submatrix

[Karol L. <kar...@kn...>]

On Wednesday 14 June 2006 11:14, Martin Wiechert wrote:
> Hi list,
>
> is there a concise way to address a subrectangle of a 2d array? So far I'm
> using
>
> A [I] [:, J]
>
> which is not pretty and more importantly only works for reading the
> subrectangle. Writing does *not* work. (Cf. session below.)
>
> Any help would be appreciated.
>
> Thanks,
> Martin

You can achieve this by using the "take" function twice, in this fashion:

>>> a =3D numpay.ones((10,10))
>>> for i in range(5):
=2E..  for j in range(5):
=2E..   a[i][j] =3D i+j
=2E..=20
>>> a
array([[0, 1, 2, 3, 4],
       [1, 2, 3, 4, 5],
       [2, 3, 4, 5, 6],
       [3, 4, 5, 6, 7],
       [4, 5, 6, 7, 8]])
>>> print a.take.__doc__
a.take(indices, axis=3DNone).  Selects the elements in indices from array a=
=20
along the given axis.
>>> a.take((1,2,3),axis=3D0)
array([[1, 2, 3, 4, 5],
       [2, 3, 4, 5, 6],
       [3, 4, 5, 6, 7]])
>>> a.take((1,2,3),axis=3D0).take((2,3),axis=3D1)
array([[3, 4],
       [4, 5],
       [5, 6]])

Cheers,
Karol

=2D-=20
written by Karol Langner
=B6ro cze 14 11:27:33 CEST 2006

Re: [Numpy-discussion] addressing a submatrix

[Martin W. <Mar...@mp...>]

Hi Simon,

thanks for your reply.

A [I, J]

seems to only work if the indices are *strides* as in your example. I need 
fancy indices (like I = (1,3,4), J = (0,3,5)), and for them A [I, J] won't do 
what I want. As you can see from the example session I posted it does not 
address the whole rectangle IxJ but only the elements (I_1, J_1), (I_2, J_2). 
E.g., if I==J this is the diagonal of the submatrix, not the full submatrix.

Martin


On Wednesday 14 June 2006 20:25, Simon Burton wrote:
> On Wed, 14 Jun 2006 11:14:17 +0200
>
> Martin Wiechert <mar...@gm...> wrote:
> > Hi list,
> >
> > is there a concise way to address a subrectangle of a 2d array? So far
> > I'm using
> >
> > A [I] [:, J]
>
> what about A[I,J] ?
>
> Simon.
>
> >>> import numpy
> >>> a=numpy.zer
>
> numpy.zeros       numpy.zeros_like
>
> >>> a=numpy.zeros([4,4])
> >>> a
>
> array([[0, 0, 0, 0],
>        [0, 0, 0, 0],
>        [0, 0, 0, 0],
>        [0, 0, 0, 0]])
>
> >>> a[2:3,2:3]=1
> >>> a
>
> array([[0, 0, 0, 0],
>        [0, 0, 0, 0],
>        [0, 0, 1, 0],
>        [0, 0, 0, 0]])
>
> >>> a[1:3,1:3]=1
> >>> a
>
> array([[0, 0, 0, 0],
>        [0, 1, 1, 0],
>        [0, 1, 1, 0],
>        [0, 0, 0, 0]])

Re: [Numpy-discussion] addressing a submatrix

[Karol L. <kar...@kn...>]

On Wednesday 14 June 2006 11:14, Martin Wiechert wrote:
> is there a concise way to address a subrectangle of a 2d array? So far I'm
> using
>
> A [I] [:, J]
>
> which is not pretty and more importantly only works for reading the
> subrectangle. Writing does *not* work. (Cf. session below.)
>
> Any help would be appreciated.
>
> Thanks,
> Martin

You can also use A[m:n,r:s] to refernce a subarray.

=46or instance:
>>> a =3D numpy.zeros((5,5))
>>> b =3D numpy.ones((3,3))
>>> a[1:4,1:4] =3D b
>>> a
array([[0, 0, 0, 0, 0],
       [0, 1, 1, 1, 0],
       [0, 1, 1, 1, 0],
       [0, 1, 1, 1, 0],
       [0, 0, 0, 0, 0]])

Cheers,
Karol

=2D-=20
written by Karol Langner
=B6ro cze 14 11:49:35 CEST 2006

Re: [Numpy-discussion] addressing a submatrix

[Pau G. <pau...@gm...>]

On 6/14/06, Martin Wiechert <Mar...@mp...> wrote:
> Hi Simon,
>
> thanks for your reply.
>
> A [I, J]
>
> seems to only work if the indices are *strides* as in your example. I need
> fancy indices (like I = (1,3,4), J = (0,3,5)), and for them A [I, J] won't do
> what I want. As you can see from the example session I posted it does not
> address the whole rectangle IxJ but only the elements (I_1, J_1), (I_2, J_2).
> E.g., if I==J this is the diagonal of the submatrix, not the full submatrix.

you can use A[ ix_(I,J) ] to do what you want.

But, if you just want subrectangular regions then A[1:4,1:4] is enough.
Please note that A[1:4,1:4] is not the same as A[ arange(1,4), arange(1,4) ],
but is the same as A[ ix_(arange(1,4), arange(1,4)) ].

hope this heps
pau

Re: [Numpy-discussion] addressing a submatrix

[Martin W. <mar...@gm...>]

Thanks Pau, that's exactly what I was looking for.

Martin


On Wednesday 14 June 2006 12:01, you wrote:
> On 6/14/06, Martin Wiechert <Mar...@mp...> wrote:
> > Hi Simon,
> >
> > thanks for your reply.
> >
> > A [I, J]
> >
> > seems to only work if the indices are *strides* as in your example. I
> > need fancy indices (like I = (1,3,4), J = (0,3,5)), and for them A [I, J]
> > won't do what I want. As you can see from the example session I posted it
> > does not address the whole rectangle IxJ but only the elements (I_1,
> > J_1), (I_2, J_2). E.g., if I==J this is the diagonal of the submatrix,
> > not the full submatrix.
>
> you can use A[ ix_(I,J) ] to do what you want.
>
> But, if you just want subrectangular regions then A[1:4,1:4] is enough.
> Please note that A[1:4,1:4] is not the same as A[ arange(1,4), arange(1,4)
> ], but is the same as A[ ix_(arange(1,4), arange(1,4)) ].
>
> hope this heps
> pau