[Symphony] Rsymphony and package 'bigmatrix'

[Marko Limbek]

Hello,

Maybe my question was too long.

Can I use object 'big.matrix' and 'filebacked.big.matrix'  in Rsymphony
package?

Is there any other way to use large matrices in Rsymphony, lets say 100.000
x 100.000 that take 50 GB of drive space? I don't know how else to work
with large matrix if there is any other way to represent a large matrix
with less space and I am a bit confused!! I am really new to such big data,
because here I think I have already entered through the door of big data!

Thanks a lot
Makro

On Wed, Jan 16, 2013 at 1:41 PM, Marko Limbek <marko.limbek at gmail.com>wrote:

> Hello everyone
>
> I am turning to the list with my newest issue :)
> I have been working on some ILP problem and I have reached the limit of
> 'matrix' function and within it 'dim' function, that allows me dimensions
> up to 10.000 x 10.000 or 15.000 x 15.000, but my current problem is 90.000
> x 90.000 at least with my current constraints construction. And I might
> encounter problem of dimension 10^6 x 10^6. Than retreat and admit defeat?
> So I have been thinking about using package 'bigmatrix', but the problem
> is, that function 'Rsymphony_solve_LP' accepts 'matrix' in the place of
> 'mat' attribute and not 'filebacked.big.matrix'. Because it takes ncol(mat)
> and ncol(row) and object of type 'big.matrix' is list and it is quite
> complicated to reach actual dimensions:
>
> Easy access to dimensions:
>
> > A <- cbind(diag(1000), diag(1000), diag(1000), diag(1000))
> > ncol(A)
> [1] 4000
>
> Complicated access to dimensions:
>
> fbm <- filebacked.big.matrix(5000,5000,
> init=0,  backingpath="path",  backingfile="bm.txt", descriptorfile="descr.txt")
> A1 <- cbind(fbm, fbm, fbm, fbm)
> > ncol(A1[1][[1]][,])
> [1] 5000
>
> Plus that Rsymphony_solve_LP uses a internal function 'make_csc_matrix',
> which I can't find.
> So I think that is quite complicated and I am looking for a way to use
> 'bigmatrix'. And also I have to make bigmatrix work and give me matrices of
> dimensions 10^5 x 10^5 and that I also haven't managed yet.
>
>
> Before I have also tried to use only 'big.matrix' but it cannot even make
> a matrix of 20.000 x 20.000
>
> > BM <-big.matrix(20000,20000, init=0, type="integer")
> Error in big.matrix(20000, 20000, init = 0, type = "integer") :
>   Error: memory could not be allocated for instance of type big.matrix
>
> So, would it perhaps be possible for me to get a small piece of advice by
> anyone, whether I should continue solving that large problem in R?
>
>
> Thank you a lot,
>
>
> Marko
>
> PS: By the way, is it also possible that I ask someone about my
> constraints construction, because maybe it would be possible to reduce it?
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/symphony/attachments/20130117/29db4008/attachment.html>