🚀 go-pugleaf

In article <erwin.725225662@trwacs>, erwin@trwacs.fp.trw.com (Harry Erwin) writes:
> Our newsmaster (me) has been learning about batching the hard way. The
> following is being reposted:
> 
>>References: <H.eg.UEnwI&3EHKA@wpcst1.phys-chemie.uni-wuerzburg.dbp.de> <1992Dec21.182154.4670@magnus.acs.ohio-state.edu> <1992Dec23.110835.77@rcwusr>
>>Date: Wed, 23 Dec 1992 20:23:15 GMT
> 
> lakerb@rcwusr.bp.com writes:
> 
>><OPINION MODE ON>
>>It seems that the chaos-focussed researchers can't see the forest for the
>>trees.  The "chaos" that is enamoured for its inherent "beauty" may, in fact,
>>be conveying a set of signals bearing important information for the control of
>>some neural process.  Decoding the "chaos" that bears the message is likely to
>>be less than fruitful  ...  the key is the range of the variation in the
>>information borne by the "chaos".
>><OPINION MODE OFF>
> 
>>Robin Lake
>>BP Research
>>lake@rcwcl1.dnet.bp.com
> 
> The chaos seen in Freeman's work appears to be associated with an
> accessible hyperbolic critical point that is used as a starting point for
> rapid pattern recognition. 

At a critical point, any of a number of successive paths (trajectories,
strategies) are equally feasible.  In that "information" is the selection from
a set of alternatives, does the pursuit of a successive path where any path is
equally possible convey any information?  If you are saying that Freeman
hypothesizes a hyperbolic critical point and assumes that the system
sensitivity at that point effects a pattern recognition, I'd be quite hesitant
to accept a no-information-value-added path as conveying any information about
the pattern presented.  It may be that the representation of the underlying
system is less than complete, neglecting to include the input representing the
pattern(s).

I'll think about this for a while.

> Another potential application of chaos is in
> controlling a significant, naturally dissipative process at minimum cost.
> A third application of chaos is in creating intelligent but unpredictable
> behavior. (A good fighter pilot is chaotic.) In yet another context, the
> "best" approach to non-zero-sum games with information collection is
> inherently chaotic (and hence difficult for purely stimulus/response
> systems to generate). Chaos can also be exploited for secure communication
> between members of a distributed community. Chaos also resists stylized
> attacks better than any periodic process (since the response time
> distribution has a continuous Fourier spectrum). 
> 

I'm willing to buy into the concept that a fractal dimension (or better yet, a
CHANGE in a fractal dimension) can convey information.  That this represents
"intelligence" requires a leap of faith.

Rob Lake
BP Research
lake@rcwcl1.dnet.bp.com