[jgap-users] JGAP progress 'stalls'

[Leander Q. <lea...@gm...>]

Hi,
I've been using JGAP for the past few months, and continue running  
into the same problem.  In my application, JGAP very quickly (<200  
evolutions with a population size of ~250) achieves fitness values of  
approximately 84% of a known problem optimal value.  However, after  
this, very little progress is made.  Increasing the population size  
to as much as 500 and/or the increasing the maximum number of  
evolutions to as much as 2000 each only achieve another 1% of the  
optimal value.

I'm using DefaultConfiguration, except that I've set  
selectFromPrevGen to 0.8 and minPopSizePercent to 90.  Each  
chromosome (hopefully this is the right term, I'm somewhat new to  
GAs) is composed of four custom genes that are essentially arrays  
whose elements sum to (before any evolution) a prescribed value.

Why might this 'stalling' of the GA's progress towards the true  
optimal value happen?  Do I have a setting wrong in the  
Configuration, or does it sound like an error in the custom gene  
code?  Am I completely in left field?  What information would be  
useful to know about my code?

Thanks very much for your advice!

Leander