[galib] best individual satisfying constraint
mwall at oculustech.com
Wed Jul 13 14:35:37 EDT 2005
On 13 Jul 2005, at 12:11, Arthur Liew wrote:
> Dear all,
> Hi, I need to calculate the best genome for an constraint optimization
> problem. However, I notice the GAlib return the best genome for
> fittest individual which may not satisfy constraint. Any idea how to
> save the best genome satisfying constraint ? Or any idea which file
> this routine is done ? Thanks !
there are a few approaches to what you describe:
1) assign value to the amount to which a solution satisfies the
constraints, and include that in the objective score. you will have
to figure out how to keep measurement of constrained-ness from
confounding your measurement of goodness and vice-versa.
2) use a representation in which constraint violations are
impossible, but you still get maximum coverage of the search space.
you will probably have to write your own mutation and crossover
3) use a repair operator that deterministically (or not) changes a
constraint-violating solution into a properly-constrained solution.
this is often done during genome evaluation, in the objective function.
and i'm sure you can imagine other variations...
More information about the galib