Never Give Up on a GA
I've actually been working with evolutionary programming techniques on my latest project. It's fun and a bit frustrating at the same time. One thing I am learning, though, is even though GAs may be slow (or my implementation really really sucks (which is very very possible)) they can make unexpected leaps. Take a look at this picture:
For a long time during the generations, it was "stuck". Then, it randomly found a new solution that was much better, and that pushed it through to giving me an acceptable solution. Sure, it took 30 minutes to get there, but if I tried to come up with the solution myself I'd still be shuffing data around!
* Posted at 08.05.2009 01:00:53 PM CST | Link *