pysparse-users

[Pysparse-users] Adding a column of a sparse matrix to a dense vector

[Artur P. <art...@gm...>]

Hi to you all, I hope someone can help me.

I am trying to do the following with pysparse but I am having problems doing it.

I have a dense vector, say:

x = numpy.arange(0,10)

and a sparse matrix:

A = pysparse.spmatrix.ll_mat(10,10)

What I wish to do is:

x = x + A[:, 2]

But this gives the following error:
TypeError: unsupported operand type(s) for +: 'float' and 'll_mat'

Then I tried like this:

x = numpy.arange(0,10)

A = pysparse.sparray.sparray((10,10))

With this it works ok but then, when I try to do a slice:

A_reduced = A[0:5, 0:5]

It gives an error:
TypeError: unsupported operand type(s) for +: 'int' and 'slice'

Anybody can give a help?

Also, what is the main difference between spmatrix and sparray?

Thanking in advance

Best regards

-artur palha

Re: [Pysparse-users] Adding a column of a sparse matrix to a dense vector

[Daniel W. <dan...@gm...>]

On Tue, Jun 17, 2008 at 5:33 AM, Artur Palha <art...@gm...> wrote:
> Hi to you all, I hope someone can help me.
>
> I am trying to do the following with pysparse but I am having problems doing it.
>
> I have a dense vector, say:
>
> x = numpy.arange(0,10)
>
> and a sparse matrix:
>
> A = pysparse.spmatrix.ll_mat(10,10)
>
> What I wish to do is:
>
> x = x + A[:, 2]
>
> But this gives the following error:
> TypeError: unsupported operand type(s) for +: 'float' and 'll_mat'

If I remember correctly, ll_mat would need to have the __iter__ method
to work with numpy. numpy checks the type of ll_mat then looks for one
of __getitem__, __iter__, __array__ or   __add__. I don't think ll_mat
has these. The code I use has a wrapper which makes pysparse more
pythonic. You may prefer to use the wrapper if it suits your purposes.

     <http://matforge.org/fipy/browser/trunk/fipy/tools/pysparseMatrix.py>


>
> Then I tried like this:
>
> x = numpy.arange(0,10)
>
> A = pysparse.sparray.sparray((10,10))
>
> With this it works ok but then, when I try to do a slice:
>
> A_reduced = A[0:5, 0:5]
>
> It gives an error:
> TypeError: unsupported operand type(s) for +: 'int' and 'slice'

Not sure about why this doesn't work. It was probably implemented and
never tested properly.  I don't believe it does anything useful
anyway.

> Anybody can give a help?
>
> Also, what is the main difference between spmatrix and sparray?

I would avoid using sparray. You also might want to check out scipy as
their sparse support is much better now.

-- 
Daniel Wheeler

Re: [Pysparse-users] Adding a column of a sparse matrix to a dense vector

[Artur P. <art...@gm...>]

Thank you for your help Daniel

I have downloaded it but it is asking for more modules from your project.

I will try to have a look and see if I can do something based on what
you did. In the meantime I will also have a look a pyTrilinos.

Thank you once again.

-artur palha

On Tue, Jun 17, 2008 at 11:33 AM, Artur Palha <art...@gm...> wrote:
> Hi to you all, I hope someone can help me.
>
> I am trying to do the following with pysparse but I am having problems doing it.
>
> I have a dense vector, say:
>
> x = numpy.arange(0,10)
>
> and a sparse matrix:
>
> A = pysparse.spmatrix.ll_mat(10,10)
>
> What I wish to do is:
>
> x = x + A[:, 2]
>
> But this gives the following error:
> TypeError: unsupported operand type(s) for +: 'float' and 'll_mat'
>
> Then I tried like this:
>
> x = numpy.arange(0,10)
>
> A = pysparse.sparray.sparray((10,10))
>
> With this it works ok but then, when I try to do a slice:
>
> A_reduced = A[0:5, 0:5]
>
> It gives an error:
> TypeError: unsupported operand type(s) for +: 'int' and 'slice'
>
> Anybody can give a help?
>
> Also, what is the main difference between spmatrix and sparray?
>
> Thanking in advance
>
> Best regards
>
> -artur palha
>

Re: [Pysparse-users] Adding a column of a sparse matrix to a dense vector

[Daniel W. <dan...@gm...>]

On Fri, Jun 20, 2008 at 6:50 AM, Artur Palha <art...@gm...> wrote:
> Thank you for your help Daniel

No problem.

> I have downloaded it but it is asking for more modules from your project.

Yes. I realized that Dominique Orban created a version of this that
works with pysparse alone. See attachment. It really needs to be
integrated into pysparse properly. I just haven't had a chance yet.

> I will try to have a look and see if I can do something based on what
> you did. In the meantime I will also have a look a pyTrilinos.

Try scipy as well.

-- 
Daniel Wheeler

Re: [Pysparse-users] Adding a column of a sparse matrix to a dense vector

[Dominique O. <dom...@gm...>]

Artur,

On top of the file that Daniel sent, you also need sparseMatrix.py
which you can download from the repository Daniel mentioned earlier. I
also attach it below.

Good luck,
Dominique

On Fri, Jun 20, 2008 at 10:09 AM, Daniel Wheeler
<dan...@gm...> wrote:
> On Fri, Jun 20, 2008 at 6:50 AM, Artur Palha <art...@gm...> wrote:
>> Thank you for your help Daniel
>
> No problem.
>
>> I have downloaded it but it is asking for more modules from your project.
>
> Yes. I realized that Dominique Orban created a version of this that
> works with pysparse alone. See attachment. It really needs to be
> integrated into pysparse properly. I just haven't had a chance yet.
>
>> I will try to have a look and see if I can do something based on what
>> you did. In the meantime I will also have a look a pyTrilinos.
>
> Try scipy as well.
>
> --
> Daniel Wheeler
>
> -------------------------------------------------------------------------
> Check out the new SourceForge.net Marketplace.
> It's the best place to buy or sell services for
> just about anything Open Source.
> http://sourceforge.net/services/buy/index.php
> _______________________________________________
> Pysparse-users mailing list
> Pys...@li...
> https://lists.sourceforge.net/lists/listinfo/pysparse-users
>
>

Re: [Pysparse-users] Adding a column of a sparse matrix to a dense vector

[gorkiana m. <gor...@gm...>]

Dominique,

Yes, I noticed that, I also made some changes because it asked numerix which
was inside Fipy. Thank you. But yet I could not use it.

Nevertheless I managed to solve my problem during this weekend.

Basically pysparse.spmatrix.ll_mat objects have two methods: keys() and
values()

With this two it is possible to do what I want. This is what I did:

A = pysparse.spmatrix.ll_mat(10,10)   # a 10 x 10 sparse matrix
X = numpy.arange(0,10)                       # a dense vector of length 10

A[0,1] = 10; A[0,2] = 10; A[0,3] = 10

To add the first row of A to X one just needs to do the following:

A_indices = numpy.array(A[0,:].keys())[:,1]     # keys is a list of tuples,
so converting to numpy.array yelds a matrix
                                                                         #
the first column is the row index and the second column the column index
                                                                         #
we just need the column index
                                                                         #
we get a vector with the indices where A has nonzero values at row 0
A_values = numpy.array(A[0,:].values())         # the nonzero values of row
0

X[A_indices] = X[A_indices] + A_values

Simple!

Thank you very much Dominique and Daniel for your help. Probably I will have
more looks at FiPy because I will have more problems for sure!

-artur palha

On Sat, Jun 21, 2008 at 2:58 PM, Dominique Orban <dom...@gm...>
wrote:

> Artur,
>
> On top of the file that Daniel sent, you also need sparseMatrix.py
> which you can download from the repository Daniel mentioned earlier. I
> also attach it below.
>
> Good luck,
> Dominique
>
> On Fri, Jun 20, 2008 at 10:09 AM, Daniel Wheeler
> <dan...@gm...> wrote:
> > On Fri, Jun 20, 2008 at 6:50 AM, Artur Palha <art...@gm...>
> wrote:
> >> Thank you for your help Daniel
> >
> > No problem.
> >
> >> I have downloaded it but it is asking for more modules from your
> project.
> >
> > Yes. I realized that Dominique Orban created a version of this that
> > works with pysparse alone. See attachment. It really needs to be
> > integrated into pysparse properly. I just haven't had a chance yet.
> >
> >> I will try to have a look and see if I can do something based on what
> >> you did. In the meantime I will also have a look a pyTrilinos.
> >
> > Try scipy as well.
> >
> > --
> > Daniel Wheeler
> >
> > -------------------------------------------------------------------------
> > Check out the new SourceForge.net Marketplace.
> > It's the best place to buy or sell services for
> > just about anything Open Source.
> > http://sourceforge.net/services/buy/index.php
> > _______________________________________________
> > Pysparse-users mailing list
> > Pys...@li...
> > https://lists.sourceforge.net/lists/listinfo/pysparse-users
> >
> >
>