[mlpack] Out of memory error

[Anisa Llaveshi]

Dear Ryan,

Thank you very much for your help. I am using the modified version now and
it scales much better.

Best regards,
Anisa Llaveshi

On Fri, Aug 24, 2018 at 4:09 AM Ryan Curtin <ryan at ratml.org> wrote:

> On Thu, Aug 23, 2018 at 03:06:48PM +0200, Anisa Llaveshi wrote:
> > Dear Ryan,
> >
> > Thank you for your quick response. I double checked and I am afraid that
> is
> > not the case. The matrix that I am providing as input to the model is a
> 1xN
> > matrix. I explicitly check the number of rows (1) and the number of
> columns
> > (N) of the matrix. The model would not even allow that because the labels
> > provided to the model is a row vector (rowvec) of N columns so the
> > dimensions would not match.
> >
> > Please let me know in case you think I am missing something or if you
> have
> > another idea of what might be going on.
>
> Hi Anisa,
>
> It turns out you aren't missing anything at all.  I took a look at the
> linear regression implementation and was surprised to see that it
> actually is trying to form an NxN matrix, whereas for linear regression
> it should only be necessary to form a dxd matrix.
>
> I spent a little time this evening reimplementing the code and have
> opened a pull request to fix it:
>
> https://github.com/mlpack/mlpack/pull/1500
>
> You can either work directly off that branch, wait for another release,
> or copy the modified linear_regression.cpp file into your codebase.
>
> But now for your problem with 200k points, it should be blazing fast
> (like 0.005s to solve the system, minus data loading time).  So you
> should be able to scale much more.
>
> Sorry for the error!  I hope that this helps out, and if not, just let
> me know.
>
> Thanks,
>
> Ryan
>
> --
> Ryan Curtin    | "The enemy cannot press a button... if you have
> ryan at ratml.org | disabled his hand." - Sgt. Zim
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://knife.lugatgt.org/pipermail/mlpack/attachments/20180824/f29f2a26/attachment.html>