RE: Friam Digest, Vol 19, Issue 10

[Nick Thompson]

Steve,

Do these models have adaptation?   Most models of this type that I have
been exposed to  have evolution and natural selection but not ADAPTATION.  

Just being annoying,

Nick

Nicholas S. Thompson
Professor of Psychology and Ethology
Clark University
[hidden email]
http://home.earthlink.net/~nickthompson/
 [hidden email]

>
>
> Today's Topics:
>
>    1. Darwin@Home (Stephen Guerin)
>
>
> ----------------------------------------------------------------------
>
> Message: 1
> Date: Mon, 10 Jan 2005 00:27:13 -0700
> From: "Stephen Guerin" <[hidden email]>
> Subject: [FRIAM] Darwin@Home
> To: "Friam" <[hidden email]>
> Message-ID: <[hidden email]>
> Content-Type: text/plain; charset="iso-8859-1"
>
> Biota.org is back up and promoting Darwin@Home.
> http://www.darwinathome.org/mission/index.html
>
> Some familiar Alife applications are adapting to it:
>
>   - Fluidium: Gerald De Jong's "tensegrity" structures. Nice example of

JOGL

> and webstart
> I believe Owen passed around a link 6 months ago:
> http://fluidiom.sourceforge.net/
>
>   - SodaBot -> SodaRace
>
>   - Larry Yaeger's PolyWorld
>
>   - More at: http://www.darwinathome.org/teams/index.html
>
> -Steve
>
> ________________________________________________________
> [hidden email] http://www.redfish.com
> office: [phone] 624 Agua Fria Street
> mobile: [phone] Santa Fe, NM 87501
>
>
>
>
> ------------------------------
>
> _______________________________________________
> Friam mailing list
> [hidden email]
> http://redfish.com/mailman/listinfo/friam_redfish.com
>
>
> End of Friam Digest, Vol 19, Issue 10
> *************************************

[Stephen Guerin]

Nick writes:
> Do these models have adaptation?   Most models of this type that I have
> been exposed to  have evolution and natural selection but not
> ADAPTATION.

Hmm, you may define "adaptation" in a more rigorous way than I.

I consider natural selection to be a form of adaptation. Learning during an
agent's lifetime is another form of adaptation. And, manipulation of the
environment, as in the indirect communication via pheromone fields in ant
foraging, to be a third form of adaptation for an agent system.

As a generality, I'd say most Alife models primary mechanism of adaptation
is natural or artificial selection.

However, Larry Yaeger's Polyworld, which is one of the models on the link,
does include non-hereditary adaptation through Hebbian learning. From his
paper (http://www.beanblossom.in.us/larryy/Yaeger.ALife3.pdf):

"PolyWorld brings together biologically motivated
genetics, simple simulated physiologies and metabolisms, Hebbian learning in
arbitrary neural network
architectures, a visual perceptive mechanism, and a suite of primitive
behaviors in artificial organisms
grounded in an ecology just complex enough to foster speciation and
inter-species competition.
Predation, mimicry, sexual reproduction, and even communication are all
supported in a
straightforward fashion. The resulting survival strategies, both individual
and group, are purely
emergent, as are the functionalities embodied in their neural network
"brains". Complex behaviors
resulting from the simulated neural activity are unpredictable, and change
as natural selection acts over
multiple generations."

Your point is interesting. I guess what constitutes an Alife model is rather
fuzzy. In the late 80s and early 90s I'd say ~70% of Alife models had GA/GP
mechanisms as central components. That said, tangentially related models
like flocking, ant foraging models and machine learning models were also
included in the conferences. Since the mid-90s, I think the meaning of what
constitutes a living system to the Alife community has pushed out from a
naive application of neo-darwinist mechanism. Some would argue for the
necessary presence of generalized thermodynamic work-cycles for an Alife
system to be considered "alive". "Some" being me with a mouse in my pocket
;-)

-S
________________________________________________________
[hidden email] http://www.redfish.com
office: [phone] 624 Agua Fria Street
mobile: [phone] Santa Fe, NM 87501

> -----Original Message-----
> From: Nicholas Thompson [mailto:[hidden email]]
> Sent: Monday, January 10, 2005 11:25 AM
> To: [hidden email]
> Subject: [FRIAM] RE: Friam Digest, Vol 19, Issue 10
>
>
> Steve,
>

>
> Just being annoying,
>
> Nick
>
> Nicholas S. Thompson
> Professor of Psychology and Ethology
> Clark University
> [hidden email]
> http://home.earthlink.net/~nickthompson/
>  [hidden email]
>
>
> >
> >
> > Today's Topics:
> >
> >    1. Darwin@Home (Stephen Guerin)
> >
> >
> > ----------------------------------------------------------------------
> >
> > Message: 1
> > Date: Mon, 10 Jan 2005 00:27:13 -0700
> > From: "Stephen Guerin" <[hidden email]>
> > Subject: [FRIAM] Darwin@Home
> > To: "Friam" <[hidden email]>
> > Message-ID: <[hidden email]>
> > Content-Type: text/plain; charset="iso-8859-1"
> >
> > Biota.org is back up and promoting Darwin@Home.
> > http://www.darwinathome.org/mission/index.html
> >
> > Some familiar Alife applications are adapting to it:
> >
> >   - Fluidium: Gerald De Jong's "tensegrity" structures. Nice example of
> JOGL
> > and webstart
> > I believe Owen passed around a link 6 months ago:
> > http://fluidiom.sourceforge.net/
> >
> >   - SodaBot -> SodaRace
> >
> >   - Larry Yaeger's PolyWorld
> >
> >   - More at: http://www.darwinathome.org/teams/index.html
> >
> > -Steve
> >
> > ________________________________________________________
> > [hidden email] http://www.redfish.com
> > office: [phone] 624 Agua Fria Street
> > mobile: [phone] Santa Fe, NM 87501
> >
> >
> >
> >
> > ------------------------------
> >
> > _______________________________________________
> > Friam mailing list
> > [hidden email]
> > http://redfish.com/mailman/listinfo/friam_redfish.com
> >
> >
> > End of Friam Digest, Vol 19, Issue 10
> > *************************************
>
>
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9AM @ Jane's Cafe
> Lecture schedule, archives, unsubscribe, etc.:
> http://www.friam.org
>
>

[Carl Tollander-2]

Nick wrote:
>>>>Do these models have adaptation?   Most models of this type that I have
>>>>been exposed to  have evolution and natural selection but not
>>>>ADAPTATION.
>>>

I'm not sure that such models can be said to "have" adaptation.
"Adaptation" as a term may be getting overloaded in the same way "Agent"
has over the last few years.  Maybe someday soon somebody will write
an article "Agents Considered Harmful".  Heh.

Anyhow.  I have a different idea of what adaptation might usefully
mean, I don't think many ALife models fit that definition, and
I think we might make some interesting kinds of progress in
applied and theoretical complexity if we consider things from a
non-agent-centric perspective.

I am slowly coming around to the idea that the neglected notion in
complexity modeling is the way in which a model (lets call it a
phenotype) is co-produced by some bag of genes and some sufficiently
interesting environment.  So the idea of "adaptation" for me is
more the process of "situating" some GRN with respect to some
world dynamics.

For many ALife models I've seen, a 1:1 mapping between
some fixed sequence or network of genes and the phenotype traits
is assumed.  All the action occurs out in the world after the
phenotype springs fully grown from the forehead of Zeus.  These
models are thus not "adaptive" in my sense of the term.  There
is some sense that the population of such agents adapts over
time, though the notion that said population somehow is itself
a phenotype is less than satisfying in a way that I am still
having some difficulty articulating.

I've been a fan of the thermodynamic-work-cycles-centric idea for
awhile, but I still don't see from a modeling standpoint any
mechanism stated so far for choosing WHICH gradients to pick
out of the genetic/environmental soups.  I suspect that the
venue to think about these things is Artificial Epigenesis (AE).
I am hopeful that this is where we can gain some traction
in composition, karyotypic plasticity, and validation and
verification issues.  It is also one road towards some
interesting social applications.  By contrast, Polyworld
seems to have begged all of the (for me) interesting questions.

There.  Complexity might not necessarily have much to do with
what we could almost call classical notions of agents, adaptation
and artificial life.  It could have a lot to do with the
thermodynamics ideas we've been discussing in a new way that
we haven't played with much yet.

And I didn't even talk about Category Theory.  Oops.

Carl

Stephen Guerin wrote:

> Nick writes:
>
>>Do these models have adaptation?   Most models of this type that I have
>>been exposed to  have evolution and natural selection but not
>>ADAPTATION.
>
>
> Hmm, you may define "adaptation" in a more rigorous way than I.
>
> I consider natural selection to be a form of adaptation. Learning during an
> agent's lifetime is another form of adaptation. And, manipulation of the
> environment, as in the indirect communication via pheromone fields in ant
> foraging, to be a third form of adaptation for an agent system.
>
> As a generality, I'd say most Alife models primary mechanism of adaptation
> is natural or artificial selection.
>
> However, Larry Yaeger's Polyworld, which is one of the models on the link,
> does include non-hereditary adaptation through Hebbian learning. From his
> paper (http://www.beanblossom.in.us/larryy/Yaeger.ALife3.pdf):
>
> "PolyWorld brings together biologically motivated
> genetics, simple simulated physiologies and metabolisms, Hebbian learning in
> arbitrary neural network
> architectures, a visual perceptive mechanism, and a suite of primitive
> behaviors in artificial organisms
> grounded in an ecology just complex enough to foster speciation and
> inter-species competition.
> Predation, mimicry, sexual reproduction, and even communication are all
> supported in a
> straightforward fashion. The resulting survival strategies, both individual
> and group, are purely
> emergent, as are the functionalities embodied in their neural network
> "brains". Complex behaviors
> resulting from the simulated neural activity are unpredictable, and change
> as natural selection acts over
> multiple generations."
>
> Your point is interesting. I guess what constitutes an Alife model is rather
> fuzzy. In the late 80s and early 90s I'd say ~70% of Alife models had GA/GP
> mechanisms as central components. That said, tangentially related models
> like flocking, ant foraging models and machine learning models were also
> included in the conferences. Since the mid-90s, I think the meaning of what
> constitutes a living system to the Alife community has pushed out from a
> naive application of neo-darwinist mechanism. Some would argue for the
> necessary presence of generalized thermodynamic work-cycles for an Alife
> system to be considered "alive". "Some" being me with a mouse in my pocket
> ;-)
>
> -S
> ________________________________________________________
> [hidden email] http://www.redfish.com
> office: [phone] 624 Agua Fria Street
> mobile: [phone] Santa Fe, NM 87501
>
>