statistics questions

[Matthew Wall]

>The best genome statistic, does it store the best genome for that
>generation or the best ever encountered?

the statistics object stores both the best for the current generation 
and the best ever.  you can control how many of the 'best ever' 
genomes get stored (the more you store, the slower the performance 
since more comparisons will take place).

>If say I get a genome gets score of 32(max score) in generation 100, and
>a different genome gets a score of 32 in generation 200,
>which genome would be stored at the end of my run?

if you have told the statistics object to keep only one 'best' 
genome, then the genome with score of 32 from generation 200 will 
_not_ replace the genome with score 32 from generation 100.

if you told the statistics object to keep more than one 'best', then 
both would get recorded, but only if they are different (based on the 
comparator that is defined for the genome).

>Also is the best genome and the genomes that are kept unchanged in a
>generation, re-evaluated in every generation.

any genome that is not modified will not be re-evaluated; evaluation 
takes place only if a change was made to the genome.  if you copy a 
genome that has been evaluated, then the copy will not be evaluated 
until it is modified.

>(win95,pentiumII,VC++6)
>
>James Birmingham
>shu96jb at reading.ac.uk


Matthew Wall
mwall at domesolutions.CAN_THE_SPAM.com