RE: [jgap-users] Does this thing even work?

[Klaus M. <me...@br...>]

Billy,

one good starting point would be the examples provided with JGAP, they are
located in package "examples". E.g. class
examples.simpleBoolean.SimpleExample.

It could be that your fitness function is not implemented appropriately. If
you could send me some of your code, I could try to figure out your problem.

Best

Klaus

> -----Original Message-----
> From: Billy Perez [mailto:bi...@dr...] 
> Sent: Saturday, May 06, 2006 10:15 AM
> To: jga...@li...
> Subject: [jgap-users] Does this thing even work?
> 
> Hello,
> 
> I've been playing with JGAP for a couple days and I'm very 
> confused.  First of all, I began working with version 2.5.  
> As I stepped thru the code, I noticed lots of very glaring 
> bugs.  For example, there were if statements whose evaluation 
> could never be true;  instances of the code comparing a 
> variable to itself, etc, etc.  It was very obvious this thing 
> is straight up broken!
> 
> I upgraded to 2.6 and some of these problems have been fixed. 
>  However, the thing overall still doesn't seem to make sense. 
>  After getting frustrated with setting my own selectors and 
> genetic operators, I decided to go with the stock 
> DefaultConfiguration and step thru that.  It still makes 
> absolutely no sense.  
> 
> Basically, when I step thru GenoType.evolve(), I notice that 
> the population size approximately doubles 
> (BestFitnessSelector makes it population size * 0.9, 
> Crossover doubles it, then Mutation chops off a small 
> percentage???).  My original pop size is 200, the pop size 
> now is around 500 something... and since I set the config to 
> keep the population at the same size, near the end of 
> evolve() JGAP simply chops off the chromosomes exceeding the 
> 200 size.  WTF?!?!?
> 
> No surprise, my population does not evolve in any sort of 
> meaningful way. Does this thing even work?  What am I not getting?
> 
> Much thanks.
> 
> P.S. Here's my understanding of how something like this 
> should just play out of the box:  
> 
> 1. random init population
> 2. select survivors based on fitness
> 3. mate survivors based on crossover rate until population is 
> at the original size 4. apply a mutation based on mutation 
> rate 5. go to step 2
> 
> JGAP, out of the box, does not seem to do anything like this. 
>  Please help!
> 
> - Billy