in reply to
Genetic Programming or breeding Perls
Whoa -- someone else hit upon the same idea that I did!
Two years ago, whilst a Sophmore (yeah, in HS -- I'm only now a Senior) I needed a science project for biology. Having previously fallen in love with Perl, and being interested in GP, I also wrote my own GP system in Perl.
Your system is considerably different from mine -- probably mostly because your implementation isn't quite true to the definition of GP as defined by Koza. Normally, GP individuals are actual program trees, with branching constructs and multiple layers. This makes crossover harder than just string manipulation -- you have to keep track of the inherent structure of the individual.
Many people writing GP in C or Java use pointers to construct the tree. Perl being what it is, I wrote a tokenizer that tokenized the syntacticly correct Perl individuals, and munged them as strings. Not as "elegant," nor as fast, but muchly fun. ;>
Anyways, I was thinking, sooner or later, of throwing the code up on CPAN -- but later is the key word. I'm using my Perl GP implentation to do some research for the Westinghouse competition, which demands a 20-page paper. Which is due October 2nd. So I'm a little busy right now.. ;> Incidentally, the paper is about making distributed GP more efficient -- and I of course wrote my own client/server GP implentation in Perl. Gotta love threaded Perl. Anyways, I'll wander through here again once I have Free Time again, and post again -- hopefully with a little more clarity and content.