My GA investigations yeilded two huge insights into ToE for me.
That "survival of the fittest" was a damnable sound byte. "Demise of the least fit" or even just "survival of the fit" would fare better against the social dawinist "evo=eugenics" crowds.
And that although a mutation and it's inverse are theoreticly identicle thermodynamicly the odds against a reversal prior to some additional mutation are astronomical. Making evolution a very one way trip.
That's an interesting challenge. To make it solid you'd need some very precise wording in its description. Fitness could drop if the fitness function changes - so it would need to be unchanging. If the fitness function were noisy, you could have fitness drop in the short term just from bad luck (this happens in 3DVCE sometimes) so you'd have to speak of statistical tendencies or rule out noisy fitness functions. If the GA/GP/EA doesn't use elitism, fitness could drop with same bad luck...
You'd need to distinguish between mean population fitness and current-best-fitness. It's actually possible to have the former decline as the latter increases in some EA's.
It's a tough thing to render unambiguous. I'd never considered it before. Definitely interesting!
My GA investigations yeilded two huge insights into ToE for me.
That "survival of the fittest" was a damnable sound byte. "Demise of the least fit" or even just "survival of the fit" would fare better against the social dawinist "evo=eugenics" crowds.
And that although a mutation and it's inverse are theoreticly identicle thermodynamicly the odds against a reversal prior to some additional mutation are astronomical. Making evolution a very one way trip.
damianpoirier 3 years ago
Yes, I'd agree with you on both counts.
kjlg74 3 years ago
hehe, it's pretty darn hard to make a GP evolve something LESS FIT eh? :-)
keep trying though. This is my standing challenge to anti-evolution creationists "make a genetic algorithm that doesn't evolve"
damianpoirier 3 years ago
That's an interesting challenge. To make it solid you'd need some very precise wording in its description. Fitness could drop if the fitness function changes - so it would need to be unchanging. If the fitness function were noisy, you could have fitness drop in the short term just from bad luck (this happens in 3DVCE sometimes) so you'd have to speak of statistical tendencies or rule out noisy fitness functions. If the GA/GP/EA doesn't use elitism, fitness could drop with same bad luck...
kjlg74 3 years ago
...(continued)...
You'd need to distinguish between mean population fitness and current-best-fitness. It's actually possible to have the former decline as the latter increases in some EA's.
It's a tough thing to render unambiguous. I'd never considered it before. Definitely interesting!
kjlg74 3 years ago