Tweeking the GA

[jaybob20]

Ok after finding this project I looked around to find out about Genetic Algorithms. I found this page:
http://cs.felk.cvut.cz/~xobitko/ga/

The interesting part is the Travelling salesman problem (GA Example (TSP)
).

That got me to thinking that it would be cool to make some changes to the way the DHEP GA does it thing. I'm not sure if this would help or hurt or if the DHEP GA has been optimized already.....

Mostly I'm just hoping to find a way my P500s to get some Knacks. Things would be more interesting, preformance would not be based entirely on CPU speed. I do realize improper settings could also hurt the preformance if the GA.

[michaelgarvie]

That's quite a nice GA page there. However it is harder than it looks to 'optimize GA parameters' given that you know nothing about the fitness landscape being searched. However with our experience evolving digital logic circuits, the current operators are pretty good.
Cheers,
Miguel

[jaybob20]

I took a quick peek at the code.

I was thinking of some command line options to do the following would be interesing:

1. Change mutation rate
2. Select mutation type
c. Select crossover type
iv. Change cross over point(s)
E. Elitism amount.

And to keep things from being to wacky, there could be a certain range that these could be adjusted to. Naturally there would be the defaults that are currently used for no options present. It would be as if a human GA is at work or we could even start another GA project to come up with the best options :)

Would this help, hurt, is anyone interested, does anybody care?

[meckano]

Rock on dude!
I have no idea what you are talking about, but if it helps it can't hurt.

...so long as I learn something woohoo!

:drums: