[Comp-neuro] From Socrates to Ptolemy
rinkus at comcast.net
rinkus at comcast.net
Mon Aug 11 15:55:24 CEST 2008
A control process can be both specific and use noise. In Earlier posts, I suggested the idea of increasing a noise input (perhaps mediated by *non-synaptic* Ach or NE) into a group of competing principal cells (PCs) in proportion to the novelty of the overall system input. Broadly, this lowers the probability that the PC with the largest summed *synaptic* input wins (see details in the earlier posts), which acts to pick new codes when the input is novel (code separation) and reactivate old codes when the input is familiar (code completion). The actual injection of the noise is a random event, but the statistics of the noise generator (i.e., the parameters of the distribution from which the event is drawn) is controlled according to a specific function. And it achieves a useful computational effect, i.e., that similar inputs map to similar codes.
-------------- Original message ----------------------
From: james bower <bower at uthscsa.edu>
> Bard, smirky (hard to believe). ;-) (is it time for a limerick
> Bard??). I am sure that 'noise' presents many opportunities,
> especially with respect to my postings - :-)
> Yes, with respect to thermal channel noise -- the amount of energy
> invested by neurons in precisely controlling spike timing both
> indicates that spike timing is important, and likely is intended to
> specifically reduce the influence of the inevitable thermal noise in
> physical systems.
> By analogy, Brownian noise is a physical reality one cannot avoid --
> so the cochlea operates just above this physical limit, but then the
> brain has internal mechanisms of hyper acuity to, in effect, increase
> its resolution. Also applies to vision, electro-reception, etc.
> But the original point I was trying to make about noise, is related to
> the fact that accumulated spike train intervals produce what looks
> like random distributions of intervals -- it is a mistake to conclude,
> however, that neuronal spike trains are "noisy", especially given
> Shannon's noisy-channel coding theorem.
> While "noise" is a convenient parameter to play with in abstract
> modeling for a whole bunch of reasons -- (see simulated annealing for
> example), its use as a fundamental mechanism in real biological system
> runs against what seems to be the predisposition of biology to
> implement very specific (and sophisticated) control processes.
> That was all I was trying to say.
> BTW, I was aware I was being a bit unfair to Bard. He is agnostic
> about whether there is "real" noise or not.
> On Aug 7, 2008, at 7:19 AM, G. Bard Ermentrout wrote:
> > John:
> > This was a smirky comment I made after 2 long weeks of back and
> > forth on noise. Basically, there was a paper in Nature Reviews
> > Neuroscience, some months ago in which the authors said noise was
> > very bad. Nathan Urban, Roberto Galan, and I wrote a TINS paper (Aug
> > 2008) on a possible functional role for noise in synchrony and Jim
> > sort of tweaked the community (as he is wont to do) which started
> > this rather long discussion. We meant by noise - any broadband
> > signal which has short correlation times and did not mean noise in
> > the sense of something to get rid of. For example, set up a simple
> > recurrent sparsely connected network (ala Brunel, Sompolinsky, ...)
> > and the synaptic inputs will have exactly this type of character. In
> > any case, as you point out, there is also real thremal noise as in
> > th opening of channels, but (forgive me if I say this incorrectly),
> > I think that Jim was pointing out that channel noise cannot play
> > that much of a role since the AP of a mammalian (and other neurons)
> > neuron is extremely well-regulated.
> > Bard
> > On Thu, 7 Aug 2008, John Hallam wrote:
> >> On Fri, 1 Aug 2008, G. Bard Ermentrout wrote:
> >>> We've established that there is no "noise" in the nervous system.
> >> I am intrigued by this. Engineers in my experience mean one of
> >> two things by "noise": either the effects of stochastic processes
> >> in the physical world which affect the system they are working
> >> with, or phenomena in the physical world unaccounted for in the
> >> model they are using. The first kind of noise clearly exists in the
> >> nervous system, since there are processes (e.g. channel gates)
> >> operating at the thermal noise level...
> >> So evidently you mean something different by "noise" -- could you
> >> elucidate, or send me a pointer to the paper where this is
> >> established, please, so that I can see how the term is used in
> >> practice by comp-neuro folk?
> >> Thanks in advance,
> >> John Hallam
> Dr. James M. Bower Ph.D.
> Professor of Computational Neuroscience
> Research Imaging Center
> University of Texas Health Science Center -
> - San Antonio
> 8403 Floyd Curl Drive
> San Antonio Texas 78284-6240
> Main Number: 210- 567-8100
> Fax: 210 567-8152
> Mobile: 210-382-0553
> CONFIDENTIAL NOTICE:
> The contents of this email and any attachments to it may be privileged
> contain privileged and confidential information. This information is
> for the viewing or use of the intended recipient. If you have received
> e-mail in error or are not the intended recipient, you are hereby
> that any disclosure, copying, distribution or use of, or the taking of
> action in reliance upon, any of the information contained in this e-
> mail, or
> any of the attachments to this e-mail, is strictly prohibited and that
> e-mail and all of the attachments to this e-mail, if any, must be
> immediately returned to the sender or destroyed and, in either case,
> e-mail and all attachments to this e-mail must be immediately deleted
> your computer without making any copies hereof and any and all hard
> made must be destroyed. If you have received this e-mail in error,
> notify the sender by e-mail immediately.
> Comp-neuro mailing list
> Comp-neuro at neuroinf.org
More information about the Comp-neuro