I am reading a bit about genetic programming. I think I have a pretty good idea on how it works but I am still just a bit confused about one thing. What kind of information would you use for the genes? For example, I want to create a genetic algorithm that can find its way through a maze, not a specific maze but generally. I get that you would let a whole bunch of these randomly create algorithms run through a few mazes, determine fitness, take the best ones and create children from these, mutate, repeat till you get a working solution. But what would the actually structure of the "creatures" look like, link list of some kind, tree structure, etc. All the articles and tutorials just kind of gloss over these, which I guess is understandable since each program will be different, but I need an example of some kind. Any links, keywords, advice, book titles or what ever you could give me would be great. Thanks, Kars

I started a thread about GP very recently in this forum and there were a couple of useful links. A good introduction to the structure and mutational operators can be found here. They use a tree structure to evolve the GPs.

My approach with my GPs is slightly different, but then I don''t have a Ph.D to back it up

IMHO, it would be quite difficult to evolve a GP that intelligently navigates a maze... It''s hard to beat A* for this particular purpose.

Cédric

Thank you, that is exactly the type of thing I was looking for.

Kars