Hi, Still learning about genetic algorithms - currently reading Mat Buckland's book (AI techniques for game progging). Going through the lunar lander example, it introduces a multifactorial fitness function, where fitness is just a linear combination of the four univariate fitnesses; components of the linear operator are judged on an ad-hoc basis. Just wondering whether this is ever extended to using a covariance structure - so the fitness function is Cov * F, where F is the vector of univariate fitnesses and Cov is a non-diagonal matrix. Any examples would be interesting! Thanks, Jim.

Please ignore this post - I had a complete brainfart last night - any moderators, please feel free to delete as I don't seem to be able to.

Thanks,
Jim.