Print

Print


Actually the reason I was looking into Angeline's module acquisition
was to introduce the idea of having a module library (what Angeline
refers to as the Genetic Library Builder or GLiB).  If I understand
correctly, this module library is how the function set gets changed.
Right?

Could a module library be added for ADFs in ECJ?  I am still new to
the ADF implementation in ECJ, so is there anything in particular I
should be aware of before adding a module library?

Thanks
Randy


On Fri, Mar 13, 2009 at 7:57 PM, Sean Luke <[log in to unmask]> wrote:
> ECJ doesn't have it per se, but it'd not be hard to add.  Note that
> Angeline's module acquisition is both a bit vague and also pretty
> representation-specific.  In the GP realm the big issue with module
> acquisition would be that it changes the function set, which would require
> some bookkeeping.
>
> ECJ's primary module facility is Automatically Defined Functions, which it
> has very good support for.  Before we go down the route of how one would add
> Angeline modules to ECJ, check to see if ADFs do the job you're looking for.
>  ECJ also doesn't do Koza-style architectural-altering operations (again,
> something that could be hacked in).  It's all a matter of what people have
> needed.  :-)
>
> I note that OpenBeagle appears to have a small Angeline-style library.
>
> Sean
>
> Randy Casstevens wrote:
>>
>> Does ECJ have facilities for doing Module Acquisition for Genetic
>> Programming like described by Angeline and Pollack?
>>
>> Thanks
>> Randy
>