Friam

FRIAM and causality

Classic

List

Threaded

6 messages Options

Marcus G. Daniels

FRIAM and causality

An HTML attachment was scrubbed...
URL: http://redfish.com/pipermail/friam_redfish.com/attachments/20071207/54bf6632/attachment.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: moz-screenshot.jpg
Type: image/jpeg
Size: 64381 bytes
Desc: not available
Url : http://redfish.com/pipermail/friam_redfish.com/attachments/20071207/54bf6632/attachment.jpg

glen ep ropella

FRIAM and causality

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Marcus G. Daniels on 12/07/2007 03:07 PM:
> You're adjusting something (a rule or a number) to make one gizmo act
> like another, except with more ceremony.

No. Adjusting a rule is entirely different from adjusting a number.
The adjustment of a number merely explores a space. A number spectrum
does specify/describe a metric. So, for example, adjusting an integer
with particular boundaries for the model, say [-10, 100] provides a
well-defined space.

The adjustment of a rule, however, doesn't _necessarily_ assume any such
space or metric, which is why we have to back off to the more flexible
word "measure".

The methods used for search and optimization are different when
exploring ill-formed "spaces" versus well-formed spaces. And it's
important to point this out in order to be clear about one's research,
especially to an audience that may not be expert in modeling and simulation.

- --
glen e. p. ropella, 971-219-3846, http://tempusdictum.com
I have the heart of a child. I keep it in a jar on my shelf. -- Robert Bloch

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHWrr3ZeB+vOTnLkoRAgNGAKCmk+jBjPSKYGCkyernH9YmvIproQCcDK3O
fLMrA8EO+T+4xhbL1GhiXjg=
=/FJN
-----END PGP SIGNATURE-----

Marcus G. Daniels

FRIAM and causality

Glen wrote:
> No. Adjusting a rule is entirely different from adjusting a number.
> The adjustment of a number merely explores a space. A number spectrum
> does specify/describe a metric. So, for example, adjusting an integer
> with particular boundaries for the model, say [-10, 100] provides a
> well-defined space.
For a fixed instruction set there's a fixed set of programs that can be
encoded in a fixed sized vector. The behaviors that such a program can
exhibit are also entirely fixed given precise initial state. General
and effective methods for global search can in fact be exactly the same
for numbers and rules: 0) create a set of starting candidates 1)
evaluate them, 2) tweak the good 3) destroy the bad, 4) go to 1.

To have good optimizations for searching number spaces (more efficient
than exhaustive grid search), then additional assumptions need to be
made, such as that the numbers come from a differentiable function or
have systematic gradients. For that matter [-10, 100] is not a well
defined space for a model because there are no units, and no given
meaning to how that range ought to relate to sensitivities in other agents.

An agent model is an assembly. If a component of the assembly is
tweaked a bit, that doesn't justify calling it a whole new model any
more than if a few parameters in the model changed a bit. It is a
versioning issue.

Marcus

Phil Henshaw-2

FRIAM and causality

Well, I'd hope it was only because I don't understand the physical
system or the model invloved, but it seems from what you're saying that
the diagram was actually just a model of an argument about a model and a
system. That makes it much harder for it to display an *interesting*
way in which a real model fails to fit a real system. Is that right?

Phil Henshaw ????.?? ? `?.????
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
680 Ft. Washington Ave
NY NY 10040
tel: 212-795-4844
e-mail: pfh at synapse9.com
explorations: www.synapse9.com

> -----Original Message-----
> From: friam-bounces at redfish.com
> [mailto:friam-bounces at redfish.com] On Behalf Of Marcus G. Daniels
> Sent: Saturday, December 08, 2007 12:24 PM
> To: The Friday Morning Applied Complexity Coffee Group
> Subject: Re: [FRIAM] FRIAM and causality
>
>
> Glen wrote:
> > No. Adjusting a rule is entirely different from adjusting
> a number.
> > The adjustment of a number merely explores a space. A
> number spectrum
> > does specify/describe a metric. So, for example, adjusting
> an integer
> > with particular boundaries for the model, say [-10, 100] provides a
> > well-defined space.
> For a fixed instruction set there's a fixed set of programs
> that can be
> encoded in a fixed sized vector. The behaviors that such a
> program can
> exhibit are also entirely fixed given precise initial state. General
> and effective methods for global search can in fact be
> exactly the same
> for numbers and rules: 0) create a set of starting candidates 1)
> evaluate them, 2) tweak the good 3) destroy the bad, 4) go to 1.
>
> To have good optimizations for searching number spaces (more
> efficient
> than exhaustive grid search), then additional assumptions need to be
> made, such as that the numbers come from a differentiable function or
> have systematic gradients. For that matter [-10, 100] is not a well
> defined space for a model because there are no units, and no given
> meaning to how that range ought to relate to sensitivities in
> other agents.
>
> An agent model is an assembly. If a component of the assembly is
> tweaked a bit, that doesn't justify calling it a whole new model any
> more than if a few parameters in the model changed a bit. It is a
> versioning issue.
>
> Marcus
>
> ============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org
>
>

glen ep ropella

FRIAM and causality

In reply to this post by Marcus G. Daniels

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Marcus G. Daniels on 12/08/2007 09:24 AM:
> For a fixed instruction set there's a fixed set of programs that can be
> encoded in a fixed sized vector. The behaviors that such a program can
> exhibit are also entirely fixed given precise initial state.

Unfortunately, most models are not built atop fixed instruction sets.
Most models are built atop a very complicated stack of abstraction
layers, which means the effective number of "instructions" and terminals
(data types as well as values) is infinite. If we were building our
models in, say, assembly, I might agree with you.

> General
> and effective methods for global search can in fact be exactly the same
> for numbers and rules: 0) create a set of starting candidates 1)
> evaluate them, 2) tweak the good 3) destroy the bad, 4) go to 1.

You're playing language games. Yes, the methods _can_ be the same in
the extreme case you lay out. But, in fact they are NOT the same in
most cases.

> For that matter [-10, 100] is not a well
> defined space for a model because there are no units, and no given
> meaning to how that range ought to relate to sensitivities in other agents.

It is a well-defined space, regardless of its lack of context. Adding
more information helps make the search more effective. But, the fact
that it's a regular set of ordered values with defined boundaries makes
it an easy space to search. So, in the absence of contextual
information, it is still easy to define an algorithm to search it.

> An agent model is an assembly. If a component of the assembly is
> tweaked a bit, that doesn't justify calling it a whole new model any
> more than if a few parameters in the model changed a bit. It is a
> versioning issue.

This is a sweeping and unjustified generalization. For example, most OO
based models are not changed by changing a single component of the
model. In most OO based models, one makes changes to an entire _class_
of components. And in that sense, it is only a versioning issue for the
source code that generates the model. It is NOT a versioning issue for
the dynamic execution of the model. It is an issue of design of
experiments. The source code is not the model, it's merely one of the
many generators for the model.

- --
glen e. p. ropella, 971-219-3846, http://tempusdictum.com
To read a newspaper is to refrain from reading something worthwhile. The
first discipline of education must therefore be to refuse resolutely to
feed the mind with canned chatter. -- Aleister Crowley

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHXVKaZeB+vOTnLkoRAk1MAJ4uVTVIykm45hBQMiA+vJPN3L7G/QCbB0WK
jFN8oUYBnrilnfDj+DeaMa4=
=nCh8
-----END PGP SIGNATURE-----

glen ep ropella

FRIAM and causality

In reply to this post by Phil Henshaw-2

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Phil Henshaw on 12/08/2007 06:45 PM:
> Well, I'd hope it was only because I don't understand the physical
> system or the model invloved, but it seems from what you're saying that
> the diagram was actually just a model of an argument about a model and a
> system. That makes it much harder for it to display an *interesting*
> way in which a real model fails to fit a real system. Is that right?

Not quite. The diagram is a prop to help the audience understand how
iterative modeling works in almost any context, including the design of
software unrelated to modeling or simulation.

It's just like any visual aid to any presentation. Can it really be
that hard to understand? It's just like any of the bazillion cartoons
we see all over the place. E.g.:

http://i.usatoday.net/weather/photos/dryzap2.jpg
http://www.vtaide.com/png/images/carbonCycle.jpg
http://www.sei.cmu.edu/ideal/ideal.gif
http://www.vtaide.com/png/images/atom.jpg

- --
glen e. p. ropella, 971-219-3846, http://tempusdictum.com
The United States is a nation of laws: badly written and randomly
enforced. -- Frank Zappa

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHXVfXZeB+vOTnLkoRAuZbAKC7fzvIgaBkwwg+oJSC/H0AhjT+xACgmESc
LuYrtWLlJPqsnD7Cf8A+iWI=
=LJII
-----END PGP SIGNATURE-----