Happy New Year! Yes, it should be breed.elite.0=2 The tutorial had
not been updated to reflect a change we had made in a recent revision
of the system. You can double-check to see that this parameter is
being used with -p print-params=true
If you're having the _best_ standardized fitness of the generation
increasing occasionally despite elitism, yes, this would imply a bug.
There have been bugs in the elitism before, so I'd be interested if you
can provide a stripped-down version for us to stare out. Also if you
can get the problem to occur on a unix box it'd be helpful.
On Dec 31, 2004, at 11:14 PM, Grant Morgan wrote:
> I tried setting elitism as instructed in the tutorial "breed.elite =
> 2", but
> on furthor searching found "breed.elite.0 = 2" and this does cause the
> object state.breeder.elite=2 in my debugger when running tutorial4
> still get raw fitness increasing on some generations (I am using
> 1.4.2_06 VM
> on a windozs machine). Does this mean elitism is not working? Is there
> a way
> to set elitism on?