[Help-glpk] Re: Feasibility Pump

[Andrew Makhorin]

> I'm using glpsol to solve sizable MIP problems (~50,000 binary
> variables after presolving), using the command line options "--tmlim
> 90 --memlim 400 --fpump --cuts --bestp --mipgap 0.05".  For problems
> of a certain size, the readout indicates that the "feasibility pump"
> kicks in relatively quickly, after which the problem is solved
> rapidly.  For problems above a certain point, however, it does not
> seem to kick in.  What can I do to improve matters?

In fpump there is no artifical limit to the number of variables. Thus,
if it cannot find a feasible solution, it may mean that the particular
instance is hard for the fpump heuristic. Could you provide the complete
terminal output for successful and unsuccessful cases? Thanks.

Andrew Makhorin