Re: [Help-glpk] Large binary integer program

[Andrew Makhorin]

> im trying to solve a large binary integer programm with about 50000 variables 
> an 6000 constraints.
> Moreover im using PyMathProg to solve the program, which makes it
> possible to use glpk in a python script and formulate a model 
> in the GNU MathProg language.

AFAIK, PyMathProg is implemented in pure Python, i.e. it is a separate
development *not* related to glpk, though it uses glpk as an underlying
lp/mip solver via PyGlpk.

> The solution process is relatively fast (~ 17s) but it takes very
> long until the solution process starts (~ 2 hours). 
> Thus the time to process the objectfunction and the constraints is
> very long.
> Is there a possibility to shorten this time or is this normal?
> Also the coefficients of the binary variables in the object function
> are very big ( ~ 10^9).
> Can this be a reason for the long duration?

I do not know PyMathProg implementation details, but if it takes two
hours to generate the model, it seems not very efficient. Did you try
using MathProg?