Thanks for the feedback Andrew and Brian. I'm also curious to see what the community thinks about this. I know
there have been a couple of similar C++ projects over the years (GALib, Evolving Objects), but I don't think that they have been terribly successful.
Sometime ago I needed to use Envolving Objects and I feel that is something discontinued (I had some compile errors, fixed and tried to report to someone and got no answers). On the other hand the field of metaheuristic search is pretty active
(academically, anyways).
You are right, this is part of the motivation for such a library :). And more, the use of metaheuristic in optimization field is also increasing. This is definitely not a summer-long task, but I think could make a nice
Boost Library, in the long run.
I think that GSoC project for starting this is a good idea because of time constraints to have something. And obviously the funds received by the student allows full time dedication. I think this is a good research area and is reasonable to explore as a
GSOC topic. I'd be interested in what sort of concepts and API emerge for such a library. I'm not sure if the topic area is one that would ultimately be included in boost or not. Strong integration with boost.mpi, boost.thread/process would make a more compelling case, though I think the concepts and API are much more important to think about for now and shouldn't be constrained by existing boost technologies.
I think that this library can have a very attractive design. I agree with you that is important to use as many Boost features as we can, but this can't be a constraint. The idea is to have a functional set of components and helper functions/templates at the end of the program for the Genetic Algorithm, built in a way to easy add/extend components for other metaheuristics. What do you think? If the community thinks that is a reasonable idea, I need a volunteer to be the mentor. Regards, -- Murilo Adriano Vasconcelos http://murilo.wordpress.com