Not about simulating nature (was Re: [GP] MEP better than GEP)
> Well, read the literature to see how the techniques haveThis is the heart of your last post. Stuff some words in
> been analysed and explored. The argument to which I am
> responding is your assertion that close
> modelling of nature is necessary.
my mouth, then argue with them.
If it will help (but it isn't likely, given your apparent
religious zeal to stomp out any discussion of biological
mechanisms) I will restate my position:
Where Biological systems address and resolve specific problems
that both biological and computational systems share, there
are likely to be mechansisms and algorithms that can be
generalized and applied to computational systems.
In that regard, I view genetic programming as a simulation of
biological genetics. To take this statement and expand it
into a call to simulatate every aspect of biological systems
is a gross insult to logic.
Anyone that knows anything about simulation will understand
the levels of models that are possible. Sure, you can refine
a model to such a complexity as to render the simulation totally
useless. And even a very gross, simple simulation will often
yield very useful results. But it is intellectually dishonest
to claim that the later observation negates any need in
the subject at hand to review the relationship between
the simulations and the system.
- Hi all,
Gordon I can agree with your statement below:
"That [survival in a constraint environment]is not a goal, it is a
tuatological outcome of the evolutionary process. Survivors, by definition,
survive --- at least until the environment changes, and they go extinct,
becoming non-survivors that by definition failed to survive."
However, I would argue that this applies to our artificial evolutionary
systems as well. The ability of an individual to meet the goal implicit in
our imposed fitness function is a stong factor in survival. However, it is
not the only factor determing an individual's (or more accurately their
offspring's) chances of survival.
The most obvious example is code growth. Put simply programs with more
'introns' are less likely to be damaged by crossover and hence their
offspring are less likely to have a fitness that is lower than their
parents, a strong survival benefit. Thus, given two individuals one of
which has a higher fitness, but fewer introns and the other of which has a
lower fitness but more introns the offspring of the second individual are
more likely to be seen in later generations. In this case survival trumps
our imposed fitness function. Other similar examples exist.
In artificial systems, just as in nature, survival is paramount. The 'goal'
imposed by the fitness function is an important factor, but not the only
factor in survival. In our evolutionary systems (just as in nature) all we
see are the survivers, who may or may not have achived our imposed goal.
Department of Computer Science
University of Idaho
Moscow, ID 82844
From: pusch@... [mailto:pusch@...]On Behalf Of
Gordon D. Pusch
Sent: Thursday, June 27, 2002 12:40 PM
Subject: Re: Not about simulating nature (was Re: [GP] MEP better than
> --- Martin Sewell <M.Sewell@...> wrote:That is not a goal, it is a tuatological outcome of the evolutionary
> > Evolution (e.g. Nature):
> > A process without a goal.
> > Stochastic process.
> > No explicit fitness function.
> Evolution has a goal, as simple as tough: survival in
> a constraint environment.
Survivors, by definition, survive --- at least until the environment
and they go extinct, becoming non-survivors that by definition failed to
There is nothing of ``purpose'' or ``goal'' in this process; it is merely a
description of what happens --- and to talk of evolution as having a
is to commit a teleological fallacy.
What we observe as ``evolution'' is simply the result of a random series of
pointless ``Red Queen's races'' --- positive feedback loops chasing their
tails, because they are all that is left after the negative feedback loops
have damped out. There is no ``goal'' to it --- it simply *is*.
-- Gordon D. Pusch
perl -e '$_ = "gdpusch\@...\n"; s/NO\.//; s/SPAM\.//; print;'
To unsubscribe from this group, send an email to:
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/