Run genetic algorithm from mps problem file

Hi,

I've got a IP scheduling problem that I would like to try solve using matlab's genetic algorithm. I'm building the problem elsewhere and only have .mps or .lp files to transfer it to MATLAB.

The genetic algorithm code doesn't accept the problem object created by the mpsread() function, stating that I need to use INTLINPROG for this problem structure. I've established that a solution is going to be virtually unattainable via simplex/B&B based methods, hence the use of a genetic algorithm.

How do I get the GA to read an MPS file?

Any help appreciated!

Cheers, Ash

Answer 1

[Alan Weiss]

If your problem is really an MILP problem, with no nonlinear constraints, then you should use intlinprog, and not a genetic algorithm. That is the truth.

You can always use mpsread to convert the problem to one that ga can use. Just edit the resulting problem structure to change the name of the solver and change the options. But don't do this if you have an MILP problem. Use intlinprog instead. If you find that intlinprog struggles with your problem, try to change some options to help it.

Alan Weiss

MATLAB mathematical toolbox documentation

[Ashley Thornton]

Thanks for the reply. I realised I was going about it the wrong way. I'm now reformulating it since the fitness function will allow me to remove a lot of variables that have to be present in a true MILP representation.