Genetic Tic-Tac-Toe Help
Okay, here's the story...
I've just completed a generic library that has a genetic algorithm to =
produce a maximally-fit "specimen". It's up and running, and works =
beautifully with the first example program I plan on releasing with =
it... one that breeds a specimen that will output a string of bits to =
match a test string.
The problem is, I want to release a second example program as well, =
something a bit more illustrative. I've written a evaluator function =
(the routine ID of the user-defined evaluator must be given to the =
library) that tests the specimens against each other in several games of =
tic-tac-toe. However, it seems that no matter how many iterations I run =
the algorithm, I wind up with MINIMALLY-fit specimens; i.e., the =
resultant "best" specimen makes nothing but illegal moves!
I don't think the problem's in the genetic library, but rather in my =
evalation function; in earlier tests, I recall getting better results, =
but then I tried improving the evalutor and can't remember what was =
different about it that allowed the positive results (the change was =
something that was needed).
I don't suppose anyone is willing to take a look at the evaluator =
function and library to see if I'm missing something obvious? You don't =
have to be an AI expert (I'm certainly not), just able to follow how the =
tic-tac-toe engine works.
Any help would be MUCH appreciated; the sooner I complete this (and =
confirm that the problem's not the library), the sooner the Euphoria =
community gets another addition to the AI collection! 
Rod Jackson
|
Not Categorized, Please Help
|
|
