Can YALMIP work with Matlab's GADS solver?

[Nazmul Islam]

I am trying to solve a mixed integer linear program with YALMIP. I am using 'lpsolve' solver in YALMIP. The program runs well for 10-15 integer variables and ~10,000 continuous variables. However, when these numbers increase (especially when number of integer variables reach 20-25), the problem takes a very long time to solve.

 

I am working in a commercial setting and I can't buy powerful solvers like CPLEX & others. I have a license for Matlab's GADS solvver, i.e., genetic algorithm toolbox. Can I use this solver with YALMIP?

 

I can try to rewrite my code so that it directly fits GADS input parameters. However, GADS requires "min f(x) s.t. A * x<= b" type of formulation. I find it difficult to generate A and b for large number and different types of variables. I prefer YALMIP's problem formulation and I just wonder if I can use GADS via YALMIP?

 

Thanks for your help in advance!

 

Nazmul

[Johan Löfberg]

No GADS is not supported. However, before you try a genetic/particle swarm/ant colony/crazy cow algorithm on a standard MILP problem, you should try alternative MILP solvers. There are some public alternatives (scip/glpk/cbc etc)

http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Category.IntegerProgrammingSolver

BTW, note that you can export the numerical model from YALMIP to extract "A" etc (export)

[Nazmul Islam]

Thanks Johan.

 

I followed http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Commands.export and used the following command:

 

[model,recoverymodel] = export(TotCon,Objective,sdpsettings('solver','lpsolve','cachesolvers',1));

model is coming out to be an empty (0x0) double.  I was able to get solution with following commands:

 

options = sdpsettings('solver','lpsolve','cachesolvers',1);

solution = solvesdp(TotCon,Objective,options);

Am I making any mistake while using 'export' command? Your feedback will be very appreciated. I can give you a more detailed example if you want.

 

 

 

 

Thanks,

 

Nazmul

 

[Johan Löfberg]

Looks like I haven't added lpsolve to the solver which can be exported directly.

Alternatively, use

solvesdp(constraints,objective,sdpsettings('savedebug',1)

and a file mxlpsolvedebug will be created before lpsolve is called. 

Alternatively, export to gurobi format or glpk format etc 

export(x>=0,x,sdpsettings('solver','gurobi'),[],[],1)

The 1 in the end is a hack to allow export despite not having the solver installed

[Nazmul Islam]

Thanks a lot for showing those commands.

 

--

 

Nazmul