[Trilinos-Users] Trilinos-Users Digest, Vol 57, Issue 17

Mark Hoemmen mhoemme at sandia.gov
Sun May 23 20:47:41 MDT 2010


On May 23, 2010, at 12:00 PM, trilinos-users-request at software.sandia.gov wrote:
> Message: 4
> Date: Sun, 23 May 2010 16:04:28 +0300
> From: "Mehmet Salih YILDIRIM" <linux at isadamlari.org>
> Subject: [Trilinos-Users] Solver Managers unconverge
> To: trilinos-users at software.sandia.gov
> Message-ID:
> 	<AANLkTimmqCnYNz1EsrpALXZubU7pHNKwgPuKx-xG67Qw at mail.gmail.com>
> Content-Type: text/plain; charset="iso-8859-1"
> 
> Hello!
> 
> Recently, we'd been trying to implement an image segmentation algorithm
> sm-ncut <http://www.cs.berkeley.edu/~malik/papers/SM-ncut.pdf> using
> trilinos. We decided to use trilinos because the implementation was supposed
> to be a distributed one. But we encountered some problems. They are:

Just as a check for your implementation, you may wish to take a look at another implementation (called "Damascene") of this algorithm:

http://www.catanzaro.name/?Software

This is an implementation for a single GPU using CUDA, so it doesn't satisfy your requirement for a distributed algorithm, but at least you can take a look at the code to see if you are forming the A and B matrices correctly.  I know the author personally and have seen him demonstrate his software live.

You might also try solving the (nongeneralized) eigenvalue problem $B^{-1} A x = \lambda x$, if your B matrix is indeed diagonal and positive.

mfh


More information about the Trilinos-Users mailing list