Yes, genetic algorithms offer no guarantees of finding the optimal solution. Usually they will find a serviceable one, often better than a hand-designed solution. However, sometimes they fail miserably and can offer a solution only slightly better than random.
It really all depends on the fitness landscape, encoding scheme, evolution parameters, etc. GAs are a black art. There is no silver bullet technique that everyone can use. You just have to get used to them and learn what generally works and what doesn't. They can be extremely useful if applied appropriately to the right problem, but it's very difficult to know how to apply them.
(机器翻译的只能关闭问题了)
英译汉