Originally posted by: SystemAdmin

[msaqib said:]

Hi everyone
I have an optimization problem that has three different decision variables. One is 2D, while the others are 3D. The constraints are, in general, nonlinear, as is the cost function. Can I solve this problem with an ILOG tool? Which one would it be? Any pointer to some examples that are similar would be very valuable.
Thanks and best regards
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[prubin said:]

Not with CPLEX (or at least not directly with CPLEX).  Variable dimension is a non-issue, but CPLEX is designed for problems with linear constraints.  Assuming that the variables are not discrete, I suspect that CP Solver would also be inappropriate.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[msaqib said:]

Thanks. Perhaps if I can piecewise linearize my constraints, I should be good to go with CPLEX on this problem, right?
Is AMPL is not part of ILOG tools? The following page claims that one can solve nonlinear optimization problems using AMPL. Is that correct? That should probably be the way to go.
http://www.ampl.com/solvers.html
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[alain.chabrier said:]

Hi,

if you want to go with multi dimensional variables and piecewise linearize the constraints, you can go with OPL-CPLEX.

You can have vars like :

dvar float+ Trans[Products][Cities][Cities];

and you have also simple piecewise syntax.

Alain
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[msaqib said:]

Awesome! Thanks Alain. I'd also like to hear about other tools for doing nonlinear constraints and multidimensional decision variables natively.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[prubin said:]

[quote author=msaqib link=topic=1226.msg3474#msg3474 date=1245823810]
Thanks. Perhaps if I can piecewise linearize my constraints, I should be good to go with CPLEX on this problem, right?


If the resulting piecewise linearization keeps the feasible region convex (and assuming the objective is linear or quadratic) then yes.

[quote author=msaqib link=topic=1226.msg3474#msg3474 date=1245823810]
Is AMPL is not part of ILOG tools? The following page claims that one can solve nonlinear optimization problems using AMPL. Is that correct? That should probably be the way to go.
http://www.ampl.com/solvers.html


AMPL is a modeling language and is not an ILOG product; ILOG is licensed to resell it.  AMPL handles rather the modeling of rather general optimization problems, but then needs to be linked to an appropriate solver (different solvers for different problem types).  For instance, for a nonconvex optimization problem, you might want to use KNITRO as the solver (with AMPL as the problem generator).

/Paul
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[msaqib said:]

prubin
Thanks a lot. That clarifies the situation a lot for me.
Best regards
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[Youngie said:]

Dear all,
I met a similar problem when using ILOG. My problem is that I defined one parameter called " Cost of Transport" as
[b]In mod. file[/b]
float CT[AssyMfgTaskList][Facility][Facility] = ...

[b]In dat. file[/b]
"//  Aa_b1
//  1      2      3    4
    [0, 0.06,  0.15,  0.35], //1
    [0.06,    0,  0.12,  0.34],//2
    [0.15, 0.12,  0,    0.47 ],//3
    [0.35,  0.34,  0.47,  0  ],//4
    ]; "
[font=Verdana][font=Verdana][u]Error Msg: Data Item "0" unexpected for "float[][{string}][{string}]"[/u][/font][/font]

Can anyone tell me how to fix this error? Thanks in advance.

Youngie
#DecisionOptimization
#OPLusingCPLEXOptimizer