[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-glpk] Re: Bug report for GLPK

From: Andrew Makhorin
Subject: [Bug-glpk] Re: Bug report for GLPK
Date: Tue, 23 Mar 2010 21:13:27 +0300

> I am a user of the glpk solver for solving linear programming
> problems. I am at present using the tool as part of my Master's thesis
> and I experienced a bug on several small example. I report one of them
> to you in the report attached to this mail.

> I would be very grateful to you if you could tell me how I can
> correct the bug, so that I can progress further in my work. For
> example, is there another version of the tool with which the problem
> would not happen? Or is there an option that I could use in order to
> make things right? 

Please note that glpkmex is a separate project. Probably you should
post your report to the bug tracker of the glpkmex project:

Looking thru your bug report I found that you perform perturbation
of the constraint matrix by adding 1e-15 to all its elements. It is
not a good idea to perturb *zero* elements in such way, because that
makes the matrix extremely badly scaled: max|aij|/min|aij| ~= 1e+15.
If the automatic scaling is enabled, there may appear large round-off
errors. I would suggest you either to disable the automatic scaling
in this case or, that is much better, not to perturb zero elements.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]