Print

Print


Richard, ECJ doesn't have many domain-specific operators -- there'd be  
just too many of them!   But they're pretty easy to implement.

Sean

On Jun 11, 2010, at 9:11 AM, Richard Otero wrote:

> Thank you for the quick response!  I have used GAs with ECJ with the
> classical operators; crossover type, crossover percentage, mutation
> rate, etc.  It's a great tool.
>
> I was wondering if ECJ had permutation-based operators as an option
> for GAs which I'd read about in a paper that used a GA for scheduling.
> I would be happy with any way to make sure the GA uses the values in
> the candidate string without replacement, that ten slots (lets say)
> have the numbers 1-10 with no repeats.
>
> I will look to learn more about ACO as well, thank you.
>
> Best regards,
> Richard
>
> On Fri, Jun 11, 2010 at 1:41 AM, Sean Luke <[log in to unmask]> wrote:
>> ECJ does GAs without any problem.  You'd have to write the fitness  
>> function
>> appropriate to the problem.
>>
>> Allow me to suggest that, from your description, a GA is not the  
>> right tool
>> for your problem.  It's a combinatorial optimization problem, and  
>> the right
>> technique for that is Ant Colony Optimization, which ECJ does not  
>> at present
>> support.
>>
>> Sean
>>
>> On Jun 11, 2010, at 1:34 AM, Richard Otero wrote:
>>
>>> I have a number of items that I am trying to find the correct  
>>> ordering
>>> for;
>>> I can determine a utility function for any combination of the unique
>>> items.
>>> A candidate solution is an ordering for the items (no item can be
>>> repeated).
>>>
>>> Does ECJ have GA functionality prepared for scheduling problems?
>>>
>>> Thank you for ECJ!
>>>
>>> Best regards,
>>> Richard
>>