[
https://issues.apache.org/jira/browse/MATH-781?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13265941#comment-13265941
]
Thomas Neidhart commented on MATH-781:
--------------------------------------
in MATH-434, floating-point comparisons have been changed to use either an:
* epsilon for any comparisons related to algorithm convergence
* ulp for any other comparisons
Now, when dropping the objective function of the first phase, the comparison is
done using ulp which is wrong imho as it is basically a convergence check. When
changing this back to an epsilon check like before, the test runs through as
expected.
> SimplexSolver gives bad results
> -------------------------------
>
> Key: MATH-781
> URL: https://issues.apache.org/jira/browse/MATH-781
> Project: Commons Math
> Issue Type: Bug
> Affects Versions: 3.0
> Environment: Windows 7 (64), jdk1.7.0_03
> Reporter: Scheiber Ernő
> Attachments: LinearProgCM.java, LinearProgCM2.java
>
>
> Methode SimplexSolver.optimeze(...) gives bad results with commons-math3-3.0
> in a simple test problem. It works well in commons-math-2.2.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
