[jira] [Resolved] (MATH-819) SimplexSolver - InfeasibleSolution when feasible

["Thomas Neidhart (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/MATH-819?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Thomas Neidhart resolved MATH-819.
----------------------------------

       Resolution: Implemented
    Fix Version/s: 3.2

In MATH-930, an additional cutOff parameter has been added to the SimplexSolver constructor in the reorganized optim package, allowing to zero-out very small numbers.

With the following values, the problem can be reliably solved:

{noformat}
new SimplexSolver(1e-6, 10, 1e-5)
{noformat}

By default the cutoff is set to 1e-12, but to get valid solutions in this case, a larger value is needed.
                
> SimplexSolver - InfeasibleSolution when feasible
> ------------------------------------------------
>
>                 Key: MATH-819
>                 URL: https://issues.apache.org/jira/browse/MATH-819
>             Project: Commons Math
>          Issue Type: Bug
>    Affects Versions: 3.1
>         Environment: Windows 7, JDK 1.7.0_03
>            Reporter: Raghu Rangarajan
>            Assignee: Thomas Neidhart
>             Fix For: 3.2
>
>         Attachments: CommonsSolver2.java, lp-octave.txt, test.log, test.mod, test.out
>
>
> I am seeing an odd behavior with the latest code in the main trunk (Directory revision: 1358535). The solver throws "NoFeasibleSolutionException" for a problem which has a feasible solution. Just by commenting out the last constraint, we get a feasible solution. And for that solution, the constraint in question does not seem to be playing a role. 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira