From newshub.ccs.yorku.ca!torn!utcsri!rpi!usc!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!jvnc.net!darwin.sura.net!cs.ucf.edu!news Thu Jul  9 16:20:02 EDT 1992
Article 6385 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!torn!utcsri!rpi!usc!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!jvnc.net!darwin.sura.net!cs.ucf.edu!news
>From: long@next1.acme.ucf.edu (Richard Long)
Subject: Re: Generalized Distributed Memory
Message-ID: <1992Jun29.141154.20922@cs.ucf.edu>
Sender: news@cs.ucf.edu (News system)
Organization: University of Central Florida
References: <650@trwacs.fp.trw.com>
Date: Mon, 29 Jun 1992 14:11:54 GMT
Lines: 41

In article <650@trwacs.fp.trw.com> erwin@trwacs.fp.trw.com (Harry Erwin)  
writes:
> Initial draft, distributed for comment.
(plausible holographic memory model for the cerebellum deleted)

Your interpretation of cerebellar structure as an interference hologram is  
interesting to me, if only because the proposed mechanism can be used for  
other purposes.  However, holographic models in general suffer from a  
severe conceptual limitation; namely, THEY ARE LINEAR MODELS.  In other  
words, the information "stored" in the hologram is unaltered (or perhaps  
degraded).  The original signal can indeed be reconstructed, more or less,  
but for what purpose?  This kind of memory is more like that of a  
computer's, in that information is stored and retrieved AS IS.  How is the  
cerebellum to know WHICH signal to reconstruct, or use such a signal once  
it is reconstructed?  

If we view the brain generally, and the cerebellum in particular, as a  
nonlinear dynamical system, we have the properties of distributed storage  
and robustness that you desire, with information processing to boot.  I  
don't think that we want our model's of memory to be bit-for-bit replay's  
of the incoming sensory information; rather, we want them to be abstracted  
and compacted representations with which we can reconstruct the signal  
rather than re-present it.  This is rather like saying that, if we are  
given the first thousand digits of Pi, what we want in our memory model is  
not a system for regurgitating those same thousand digits, but a SIMPLE  
(or simpler) algorithm for producing them all (or in the case of memory,  
the interesting ones ;^) ), plus a few more.  The algorithm captures the  
"essence" of Pi in a way that the thousand digits does not.

However, I do think that the inhibitory reconstructive processes which you  
use in your holographic model can also be useful in a nonlinear dynamical  
systems version.  BTW, I am currently reading Pribram's book.  His (and  
colleagues) derivation of a "neural wave equation" is particularly  
interesting.
--
Richard Long
Institute for Simulation and Training
University of Central Florida
12424 Research Parkway, Suite 300, Orlando, FL 32826
(407)658-5026, FAX: (407)658-5059
long@acme.ucf.edu


