Eigenvalue decomposition of a large sparse matrix

[kasme1]

I have been trying to compute the eig decomposition of 13000 x 13000 sparse matrix D using matlab command eigs (single (D)), but I get 'out of error' although memory wasn't fully utilized (with a 4GB RAM) , the help I saw online suggested using numerical methods like ARPACK (Fotran77 software) or ARPACK++ (associated with C/C++), I have no knowledge on how to use this software. Please any detailed tips are welcome. Thanks

 

[MacGyverS2000]

The MATLAB forum?

Dan - Owner

http://www.Hi-TecDesigns.com

 

[IRstuff]

Is this for school?

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[kasme1]

To MacGyverS2000; I am presently using Matlab, but I want to use C++ or C (arpack lib/lapack) to solve eigen decomposition of the above matrix bcos Matlab cannot decompose such large and sparse matrix but I dont know to use it cos I am familiar with C.
To IRstuff; it is not for school
Please any help will be appreciated

 

[FeX32]

Use singular value decomposition in Matlab. svd(M)
I recommend solving your problem in Matlab. C/C++ may also have issues with a large matrix.

To do it in C++, unless you want to write it from scratch, you will need an add on like OpenCV.


Fe

 

[kasme1]

Thanks for your response, I still got the same error when I used svd. I haven't used C++ before, but I suppose there should be a way to only transfer the matrix variable D into C++ without having to rewrite the whole programme?. meanwhile I google opencv to know what it does.

 

[IRstuff]

Sounds like you should talk to Mathworks, since it's their program.

Also, if your program is not is sufficiently efficient, it can easily blow past your memory limits, even though the array can only occupy a maximum of 1.3 GB.

You should also verify that all suggestions given in the Matlab help file have been implemented.

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize