Re:What problems have you solved using genetic algorithms/ge...
by _delirium on Saturday September 11, 2010 @01:34PM
(#1681916)
Yeah, I see them as just one tool in the randomized-optimization toolbox. It's not entirely clear to me why many people see them as the only or the first tool (perhaps the appealing biology metaphor?). In particular, I tend to like a complexity ramp: first try the dumbest thing, like random instances, then move on to randomized hill climbing or a variant like simulated annealing, then add more complex stuff on top of that, whether via GA crossover, or something that learns/exploits a statistical distribution, like MIMIC, if the simpler methods didn't work and you need to exploit the structure of the space. And with purely discrete things, I also like considering whether completely different approaches, like constraint programming, would work instead (instead of optimizing a numerical fitness/cost function, you give a list of things that must be true about any desired solution, and ask it to find one meeting the constraints).
[ Parent
| on HN ]
There are no comments on this comment yet.