Wednesday, March 17, 2004

Genetic Algorithm Optimizers

For one of my classes we have to write a genetic algorithm to optimize several problems and it's very interesting. The inherent randomness means that the answer and how long it takes to get the answer varies greatly for different runs of the same problem. This means that you can never really trust that the answer you get will be the best. The standard genetic algorithm also seems horribly inefficient to me. Because you know the characteristics of each population so it seems to me that you should ensure that you use this information as much as possible, and I don't think the standard algorithm does that. The biggest problem with a genetic algorithm is the inefficiency, and I think it could be made more efficient. I'm going to try out some of my ideas and see if they make it more efficient. I'll report.

No comments: