home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!paladin.american.edu!darwin.sura.net!wupost!sdd.hp.com!zaphod.mps.ohio-state.edu!cis.ohio-state.edu!news.sei.cmu.edu!fs7.ece.cmu.edu!crabapple.srv.cs.cmu.edu!news
- From: sef@sef-pmax.slisp.cs.cmu.edu
- Subject: Re: Neural Nets and Brains
- Message-ID: <1992Jul24.173903.175041@cs.cmu.edu>
- Date: Fri, 24 Jul 92 17:39:03 GMT
- Organization: School of Computer Science, Carnegie Mellon
- Nntp-Posting-Host: sef-pmax.slisp.cs.cmu.edu
- Lines: 41
-
-
- From: arms@cs.UAlberta.CA (Bill Armstrong)
-
- My point
- was that BP nets use continuous signals and the brain doesn't -- an
- obvious very significant difference. I was asking why people would
- expect to understand the brain by studying a system (BP) that is
- *different* at the most basic level of signalling.
-
- Well, your basic premise here is far from obvious and (in my
- never-very-humble opinion) probably wrong. Yes, neurons seem to do most of
- their long-distance signalling with discrete pulses, though as others have
- pointed out, there are also local connections that seem to use continuous
- analog signals of an electrical or chemical nature. However, even if we
- concentrate on the pulse-coded communication, it appears that most of the
- information is being carried by pulse-rate encoding -- an analog signal
- that happens to be encoded as a train of pulses rather than a DC level. By
- integrating these pulse trains, you get the old familiar
- sum-of-weighted-inputs used by BP and related models. Look for concurrency
- in the pulse trains (or introduce some other nonlinearity before the
- summation) and you get a quick and dirty probabilistic multiplication of
- two input signals.
-
- There may well be some insight to be gained by modeling these systems paths
- at the pulse, ion, and time-constant level, but if you want to move to a
- more abstract representation, something like BP (or a related continuous
- model, like sigma-pi units) seems a much better fit than Boolean logic.
- Boolean logic nets fall out of BP as a special case, but I've seen no
- reason to believe that the brain uses such special cases exclusively, or
- even favors them.
-
- -- Scott
- ===========================================================================
- Scott E. Fahlman
- School of Computer Science
- Carnegie Mellon University
- 5000 Forbes Avenue
- Pittsburgh, PA 15213
-
- Internet: sef+@cs.cmu.edu
-
-