[EvA] Trouble with constraints

[Johannes]

Hi Marcel,

thanks, for some of the variables hard constraints are indeed needed.
For some of the others, the "soft" ones are better.
Random was chosen purely for testing, with this solver I got the "value out
of bounds exception" for my equation the quickest :)

Regards

Johannes

2012/4/26 M.Kronfeld <makron at gmx.de>:
> Hi there,
>
> in EvA2, the prime mechanism for such a case would be to implement an
> own problem range. Just override
>        double getRangeLowerBound(int dim)
> and
>        double getRangeUpperBound(int dim)
> within your problem class to return the desired min/max values (e.g 0 or
> 1 in the first two dimensions and 0/max_value above). Doing that,
> negative values will not even be regarded in the first place.
>
> If you use the constraint mechanism, you are probably using soft
> constraints by default, meaning that negative values are allowed, but
> positive values are preferred with a certain weight (which can be
> parameterized somehow I assume).
>
> And by the way, I would not use random search for any plausible goal
> except proving that random search wont work ;)
>
> Best regards,
>  Marcel
>
>
> On 04/25/2012 06:02 PM, Johannes wrote:
>> > I want to solve a rather simple problem with the EvA2 toolbox
>> > using the Java interface.
>> > Unfortunately I cannot get the constraints to work as they should,
>> > I set them up, but they do not seem to be respected by the solver.
>> >
>> > If I did not misunderstand something fundamental on the issue, they
>> > are supposed to work like shown below. Still, I get results containing
>> > negative values, which should not be the case.
>> > Any hints are appreciated.
>> >
>> > Best regards
>> >
>> > Johannes
>> >
>> > I run this:
>> >
>> > double[] lösung =
>> > OptimizerFactory.optimizeToDouble(OptimizerFactory.RANDOM, new
>> > EvA2Problem10(zielfunktion));
>> >
>> > with this class:
>> >
>> > public abstract class EvA2Problem extends AbstractProblemDouble {
>> >
>> >     protected static final long serialVersionUID = 7005447722364059112L;
>> >     protected Abweichung abweichung;
>> >
>> >     public EvA2Problem(Abweichung abweichung)
>> >     {
>> >             this.abweichung = abweichung;
>> >             setWithConstraints(true);
>> >             setDefaultRange(10.0);
>> >             setConstraints(new AbstractConstraint[]{
>> >                             new IntervalConstraint(0, 0.0, 1.0),
>> >                             new IntervalConstraint(1, 0.0, 1.0),
>> >                             new IntervalConstraint(2, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(3, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(4, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(5, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(6, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(7, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(8, 0.0, Double.MAX_VALUE),
>> >                             new IntervalConstraint(9, 0.0, Double.MAX_VALUE)
>> >             });
>> >     }
>> >
>> >     @Override
>> >     public double[] eval(double[] values) {
>> >             return new double[]{abweichung.value(values)};
>> >     }
>> > }
>> > _______________________________________________
>> > EvA mailing list
>> > EvA at listserv.uni-tuebingen.de
>> > https://listserv.uni-tuebingen.de/mailman/listinfo/eva
>> >
> _______________________________________________
> EvA mailing list
> EvA at listserv.uni-tuebingen.de
> https://listserv.uni-tuebingen.de/mailman/listinfo/eva