News for hackers. Stuff that matters.
Re:HTML5 Genetic Cars by darkmighty on Tuesday June 25, 2013 @08:10PM (#5943304)
Also, I found these algorithms could always go more and more meta. For instance, in some situations a large "clone fraction" was optimal, or large "mutation rate", whilst in other cases it brought chaos and degraded the population.

Successful parameters could also become more immune to mutation with time, a nice improvement along with "speciation" you cited.

Re:HTML5 Genetic Cars by Xcelerate on Tuesday June 25, 2013 @08:18PM (#5943330)
> Also, I found these algorithms could always go more and more meta.

Haha, I've had this same thought before. I wrote a genetic algorithm to work on the sphere packing problem (in a periodic box) and I kept thinking about how I could mutate the settings within the program and optimize them, and then mutate the rate at which those settings themselves changed, and so on. I wonder if there's an optimal point to stop being "meta"?

Re:HTML5 Genetic Cars by darkmighty on Tuesday June 25, 2013 @10:22PM (#5943760)
There are information theoretic aspects behind this. You're essentially searching the design space for the fittest and want a fast way to get your answer -- this 'fastest way' is related to the way that 'grants' maximum information.

I say this because this reminded me of a wonderful (and acessible) discussion I once glanced over on the following book: http://www.cs.toronto.edu/~mackay/itprnn/book.pdf (Page 269, pdf 281) -- it offers answers to this sort of question :)

Re:HTML5 Genetic Cars by randomsearch on Tuesday June 25, 2013 @11:46PM (#5944069)
You're looking for "hyper-heuristics."

So called, because unfortunately GAs etc. had already taken the name "meta-heuristics".

Re:HTML5 Genetic Cars by tripzilch on Monday July 01, 2013 @11:52AM (#5971427)
> I wonder if there's an optimal point to stop being "meta"?

Perhaps this will help your thinking in the right direction: http://en.wikipedia.org/wiki/No_free_lunch_in_search_and_opt...

Re:HTML5 Genetic Cars by gknoy on Tuesday June 25, 2013 @09:05PM (#5943488)
We could probably find it with some kind of Monte-Carlo technique...
Re:HTML5 Genetic Cars by rmc on Wednesday June 26, 2013 @11:18AM (#5946585)
Successful parameters could also become more immune to mutation with time

I believe this happens in nature. DNA has lots of error correcting stuff to reduce errors.