[Top][All Lists]

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

Re: [Help-glpk] Asking help on an issue of solving via glpk

From: Chris Matrakidis
Subject: Re: [Help-glpk] Asking help on an issue of solving via glpk
Date: Tue, 9 Feb 2016 01:08:08 +0200

Should I suppose that the answer to my question is negative, i.e. it is not possible to force GLPK to start searching from a particular point?

Since you are using MathProg, you can duplicate the objective function as a constraint and specify the value you obtained with your heuristic as an upper limit (I'm assuming you are minimising). This way the search will not investigate nodes with higher objective values, saving time, effectively duplicating what would happen if the solver found the solution on its own. However this may make each iteration slower, depending on the structure of your problem. Also note that the solver output will not indicate the correct gap until a new solution is found. 

Nevertheless, I mast say that I have no idea about what --cuts, --pcost, --proxy, --mir and --mipgap 0.2 mean or stand for.
Could any one inform me in a few words about the meaning of all these? Or better, is there any text explaining all these?

 Some further info in addition to what Noli sent:
--pcost enables "pseudocost branching with strong branching initialisation". A google search with the quoted sentence will find lots of additional information.
--cuts enables all the implemented cut families, so --mir is redundant. Cuts are constraints added during the search that remove non integer solutions (like the current lp relaxation). 
--mipgap 0.2 instructs the solver to stop when the best solution found is within 20% of the lower (or upper) limit. Therefore the solution is certain to be at most 20% off the optimum.

Best Regards,

Chris Matrakidis

reply via email to

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