From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Tue Apr 03 2001 - 08:10:53 MDT
---------- Forwarded message ----------
Date: Tue, 3 Apr 2001 10:01:29 -0400 (EDT)
From: Randal Koene <randalk@marina.psych.mcgill.ca>
Reply-To: murg@yahoogroups.com
To: murg@yahoogroups.com
Subject: Re: [MURG] Re: "analog computer" = useless hypothesis? (fwd)
Hi Eugene,
On Tue, 3 Apr 2001, Eugene Leitl wrote:
> There's evidence(*) to suggest the barn owl uses a neat Hebbian learning
> trick to strengthen those signals in phase, and weaken those that aren't -
> and of course tons of neurons to average together.
Indeed, surprisingly it deals quite well with noise... and sometimes even
*uses* noise to its advantage. I have only a few references available
right now, but I rememeber there were more examples.
Anderson, J.S. and Lampl, I. and Gillespie, D.C. and Ferster, D., 2000,
The Contribution of Noise to Contrast Invariance of Orientation Tuning in
Cat Visual Cortex Science, 290(5498), 1968-1972
Hidaka, I. and Nozaki, D. and Yamamoto, Y., 2000, Functional Stochastic
Resonance in the Human Brain: Noise Induced Sensitization of Baroreflex
System, Physical Review Letters, 85:17, 3740-3743
Manwani, A. and Koch, C., 2001, Detecting and Estimating Signals over
Noisy and Unreliable Synapses: Information-Theoretic Analysis Neural
Computation, 13(1), 1-33
> What I really like about this is that it underlines how hard the brain has
> to work just to get a little accuracy over all that noise. An awful lot of
> those neurons are just voices in the crowd in various population encoding
> schemes. My guess is you could do the same as what the brain currently does
> with 2 or 3 magnitudes fewer neurons if they were rate based and reliable.
> That's before you even begin to consider the huge speed advantages
> artificial neurons have.
Okay, that is probably not quite true... I mean the emphasis on the
unreliability there. It is true that a single synapse is quite
unreliable. But keep in mind that (a) there are often multiple synaptic
connections, (b) many neurons can enter a mode in which they communicate
through bursts which have MUCH more reliable transmission probabilities
(e.g. during slow-wave sleep transfer of memories), (c) many neuronal
processes take place in cycles and signal transmission therefore has
multiple opportunities for success (e.g. during waking acquisition of
memories in short-term memory). The brain uses many schemes to cope with
the noise, and many of those schemes are already integrated with other
mechanisms that have to go on anyway (e.g. potentiation). Still, perfect
control of when a signal is or is not transmitted would of course be a
benefit. :)
Cheers,
Randal
_______________________________________________________________________
RANDAL A. KOENE
Neural Modeling Lab, Department of Psychology - McGill University
randalk@marina.psych.mcgill.ca, (514)-398-4319
minduploading.org, www.psych.mcgill.ca/perpg/stds/rk/
_______________________________________________________________________
------------------------ Yahoo! Groups Sponsor ---------------------~-~>
Join Audio Book Club today and receive 4 Audio Books
for 1 cent PLUS a Free Cassette Player. Choose from hundreds of the latest audio books, everything from John Grisham to Sydney Sheldon.
http://us.click.yahoo.com/.mMGlC/tfYCAA/rZ8EAA/_7uVlB/TM
---------------------------------------------------------------------_->
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:06:49 MST