You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@commons.apache.org by "Jurgen Tas (JIRA)" <ji...@apache.org> on 2010/03/31 10:42:27 UTC
[jira] Commented: (MATH-351) SimplexSolver fails to solve feasible
problem instance
[ https://issues.apache.org/jira/browse/MATH-351?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12851777#action_12851777 ]
Jurgen Tas commented on MATH-351:
---------------------------------
To put it simply, you have defined a problem of the following form:
max (0*x1 + 0*x2 + 0*x3 +...) subject to the following constraints:
x1 + x2 + ... >= some number
x2 + x3 + ... >= some number
etc...(you have defined inequalities of a similar form)
x1>=0, x2>=0,....
Ofcourse such a problem is unbounded; i.e. the solution is x1 = infinitiy, x2 = infinity,.... etc..
Maybe I don't understand the problem well, but to me it seems that the Excel program gives a wrong answer.
Jurgen
> SimplexSolver fails to solve feasible problem instance
> -------------------------------------------------------
>
> Key: MATH-351
> URL: https://issues.apache.org/jira/browse/MATH-351
> Project: Commons Math
> Issue Type: Bug
> Affects Versions: 2.0
> Environment: Windows Vista Home Premium Version 6.0 Service Pack 1, Build 6001
> Reporter: Mark Thomas
> Fix For: 2.1
>
> Attachments: SimplexFail.xlsx, TestSimplexFail.java
>
>
> SimplexSolver throws an UnboundedSolutionException on a problem instance I can optimally solve with Excel's Solver. I've kept the parameters between the two programs the same as far as I can tell (i.e. both have a precision/epsilon value of 1e-6 and a maxIterations value of 1000). I will attach a JUnit test with an example problem on which SimplexSolver fails. I will also attach an Excel spreadsheet wtih the same data and successful Solver setup in place.
> I don't know a whole lot about linear programming or Simplex, but the problem I'm attempting to solve does appear to have a fairly sparse coefficient matrix, which may be part of the problem.
> It's surprisingly difficult to find a Java-based linear programming library, so I was ecstatic when I found this. Let me know how I can help!
> Thanks!
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.