>It seems to me that the real difficulty in the present discussion has to
>do with the imprecision of the use of the word "limitation". How limited
>are morphological changes? Until this is clearly stated, there can be no
>clearly stated refutation.
>
For example, Stephen in responding to Glenn's post which mentions the fact
that search processes on a fitness landscape are not likely to find
improvements in the fitness function when they get near an optimum point,
says that _that_ effect is a demonstration that there are limits to
variation. Stephen is correct, _IF_ both of the following are true: a) The
fitness function(environment) does not change; b) Either the fitness
function is unimodal or "long jumps" in gene space are not permitted. In
real life (as seen by Kauffman anyway) neither of these conditions is true.
Real search spaces are always multimodal, fitness functions do change, and
long jumps in gene space are permitted. These are assumptions that are
sufficiently well-known in the community that studies these processes, that
they are generally assumed and discussion of them is not considered
necessary. Imagine if I had to state all the axioms of mathematics before
I could be allowed to speak or write on my current research! GM's secrets
would be safe :-). Perhaps it's time to begin gathering some definitions
into a FAQ.
Bill Hamilton | Vehicle Systems Research
GM R&D Center | Warren, MI 48090-9055
810 986 1474 (voice) | 810 986 3003 (FAX)