[ 
https://issues.apache.org/jira/browse/MATH-246?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12830716#action_12830716
 ] 

Jurgen Tas commented on MATH-246:
---------------------------------

Benjamin thank you for your reaction. However, I am not entirely sure about the 
use of this epsilon. What is e.g. the relationship  with the status of the 
system (i.e.ill-conditioning of the constraints matrix) of constraints and the 
value of epsilon? A small test I did this week was the following:

max (x1 + x2) subject to the following constraints:

x1 + x2 = 1
(1+delta)x1 + (1-delta)x2 = 1
x1, x2 >= 0

with delta a small number. The matrix A = [ (1,1), (1+eps, 1-eps)] is 
ill-conditioned. The epsilon in the simplex algorithm was set on the the 
default value of 1e-6. I found that for values of delta larger than 1e-6 the 
simplex algorithm provided an answer that does not obey all the constraints. If 
delta was smaller than epsilon the solution obeyed all the constraints. 

Any idea what this effect is, and how I can detect this for a general problem?

Regards,

Jurgen

> Simplex Method Implementation
> -----------------------------
>
>                 Key: MATH-246
>                 URL: https://issues.apache.org/jira/browse/MATH-246
>             Project: Commons Math
>          Issue Type: New Feature
>            Reporter: Benjamin McCann
>            Assignee: Luc Maisonobe
>             Fix For: 2.0
>
>         Attachments: newfiles.zip, simplex.patch, SimplexSolverTest.patch, 
> SimplexTableau.patch, test.patch
>
>
> I've created an implementation of the Simplex algorithm for optimizing 
> systems of constrained linear equations that I'd like to contribute.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to