help-glpk
[Top][All Lists]
Advanced

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

Re: [Help-glpk] MIP stopping


From: Klas Markström
Subject: Re: [Help-glpk] MIP stopping
Date: Wed, 18 May 2005 22:21:23 +0100

Hi Brady,
yes I am aware that it only applies under somewhat special circumstances. However when it applies it could make a big difference in running time and thus be worth quite a lot. If one is doing integer programing and have an objective function with integer coefficients then this should be applicable. If one wants to generalise one could also do this when the variables are integers and the coefficients are rational, since one can then in principle just multiply the objective function with the largest denominator present. However for large denominators this could lead to very large integers in the objective function.


/Klas

Klas,

I think it makes sense to add the feature you mention; I don't think
GLPK has it already, but I haven't checked the code for it.

Realize, however, that your situation is a special case in which all
variables that appear in the objective are integer variables and all of
their objective coefficients are integers as well.  The rule you
describe does not apply for general integer programs (even pure integer
programs).

Brady

Klas Markström wrote:
 Hi!
 I have been using glpsol to solve some
 minimisation integer programming problems. Right
 now the solver often finds the  optimal integer
 solution with value opt fairly fast and then uses
 a lot of time to improve the lower bound from opt
 -1 to opt. Since the solver "knows" that this is
 an integer programming problem it should be able
 to stop as soon as it has improved the lower
 bound to anything strictly above the current
 optimum minus one.

 Is this going to be changed in some upcoming
 version of glpsol or is there some simple way to
 change the code so it will stop as soon as it
 knows that the current best is the optimum?

 /Klas

 --

 ==========================================================================

 Klas Markström                email: address@hidden
 Department of Mathematics        fax:   (+46)90 786 52 22
 Umeå University                         phone: (+46)90 786 97 21
 S-901 87 Umea, Sweden

 URL: http://abel.math.umu.se/~klasm/

 ==========================================================================


--
Brady Hunsaker
Assistant Professor
Industrial Engineering
University of Pittsburgh
http://www.engr.pitt.edu/hunsaker/


--

==========================================================================

Klas Markström                 email: address@hidden
Department of Mathematics        fax:   (+46)90 786 52 22
Umeå University                         phone: (+46)90 786 97 21
S-901 87 Umea, Sweden

URL: http://abel.math.umu.se/~klasm/

==========================================================================




reply via email to

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