Re: Memory issue in SVD
- To: mathgroup at smc.vnet.net
- Subject: [mg75025] Re: Memory issue in SVD
- From: "Raj" <rajanikanth at gmail.com>
- Date: Sat, 14 Apr 2007 01:13:16 -0400 (EDT)
- References: <evn875$qs$1@smc.vnet.net>
hi!
I am not an expert on this but this may help:
Try using the SparseArray function..so instead of assigning the table
to M, try doing this
{U,S,V}=SingularValueDecomposition[SparseArray[Table[Random[],{5000},
{5000}]]];
Even then, you need atleast 3.5 GB of memory or more.
Although not related to this..you may speed up your computation by
enabling the NUMBER_OF_PROCESSORS environment variable(but you still
need good amount of memory!!!).
Good Luck,
Raj
On Apr 12, 11:33 pm, Tulga Ersal <ter... at umich.edu> wrote:
> Dear group,
>
> I apologize in advance to bother you with a topic that came up many
> times in the past, but nothing I found in the archives really helped.
>
> I am trying to find the singular value decomposition (SVD) of a big
> matrix, but Mathematica gives the following message:
>
> "No more memory available.
> Mathematica kernel has shut down.
> Try quitting other applications and then retry."
>
> As an example, create a 5000x5000 matrix with random entries and try
> to find its SVD:
>
> M=Table[Random[],{5000},{5000}];
> {U,S,V}=SingularValueDecomposition[M];
>
> My computer cannot perform the SVD and gives the message above almost
> immediately. Originally I thought the matrix was too big and my
> computer didn't really have the necessary resources, but I can
> compute the SVD using another program, so I know it is not really an
> hardware issue. I'm using Mathematica 5.2 on a PC running Windows XP,
> with 2GB RAM, 160GB hard disk, two 3.4 GHz processors.
>
> Is there a setting in Mathematica that needs tweaking to avoid the
> memory problem? I have tried setting $HistoryLength=0, or using
> Share, and neither helped. I'd appreciate any recommendations.
>
> Thank you,
> Tulga