[Clp] Using Interior Point can cause infinite loop

[John Forrest]

I replied offline, but thought I might as well say what fix was.

The Matlab interface had scaling enabled for Interior Point - this is
not recommended for normal use.

John Forrest


On Wed, 2011-10-05 at 10:38 -0400, Browne, Alexander C (Colin) B3REI
wrote:
> Hi,
>  
> I am relatively new to the CLP project.  I am using the interior point
> solver to solve a complex quadratic problem.  I am running the solver
> through Matlab, and the inbuilt solver (quadprog) does provide a
> solution to my problem (very slow).
>  
> I have encountered a bug in the code where the process will go into an
> infinte loop.  
>  
> The problem can be tracked back to
> ClpPredictorCorrector.updateSolution where the diagonal_ vector is
> been set.  Some of the elements on the vector are been set to
> infinite.  This causes the code in ClpPredictorCorrector.solve (line
> 504) to go into an infinite loop.
>  
> Currently I have modified the updateSolution function to check the
> diagonal_ vectors values is appropiate (not infinite). If it is
> infinite, then using the FUDGE functionality, set the value to
> diagonalLimit.  
>  
> This is not an ideal solution, as the CLP will return an unbounded
> solution.
>  
> Colin
>  
>  
>  
> ------------------------------------------------------------------------------
> CONFIDENTIALITY NOTICE: If you have received this email in error,
> please immediately notify the sender by e-mail at the address shown.  
> This email transmission may contain confidential information.  This 
> information is intended only for the use of the individual(s) or
> entity to 
> whom it is intended even if addressed incorrectly.  Please delete it
> from 
> your files if you are not the intended recipient.  Thank you for your 
> compliance.  Copyright (c) 2011 CIGNA
> ============================================================================== 
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp