|Just another Perl shrine
Re: Re: Re: A Beginning Guide To Evolutionary Algorithmsby biosysadmin (Deacon)
|on Oct 15, 2003 at 12:44 UTC
I don't think I said exactly what I meant in my first reply. I didn't mean that each individual mutation must increase fitness for an individual for a population. Here's a more specific explanation:
There are some sets of parameters that perform well when combined together in the same individual. One hypothesis of evolutionary algorithms is that combination of these good "building blocks" will form an overall more fit individual. There exist problems that violate this hypothesis, and are therefore very difficult for a simple genetic algorithm.
If you do a google search for the "Minimal Deceptive Problem," you should find some references for this. In general, these deceptive problems are very rare and are also easy to spot if you track the evolution of your population over time.
Hope this cleared my comment up. Again, awesome guide. :)