home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!gatech!destroyer!ubc-cs!unixg.ubc.ca!kakwa.ucs.ualberta.ca!alberta!arms
- From: arms@cs.UAlberta.CA (Bill Armstrong)
- Subject: Re: Neural Nets and Brains
- Message-ID: <arms.712351174@spedden>
- Sender: news@cs.UAlberta.CA (News Administrator)
- Nntp-Posting-Host: spedden.cs.ualberta.ca
- Organization: University of Alberta, Edmonton, Canada
- References: <arms.711986585@spedden> <1992Jul24.164544.11876@cs.ucf.edu> <arms.712027061@spedden> <1992Jul28.163206.22309@src.umd.edu>
- Date: Tue, 28 Jul 1992 19:19:34 GMT
- Lines: 42
-
- tedwards@src.umd.edu (Thomas Grant Edwards) writes:
-
- >In article <arms.712027061@spedden> arms@cs.UAlberta.CA (Bill Armstrong) writes:
- >>Thank you. A nice explanation. Unfortunately, taking phase and
- >>density of pulses into account makes for quite a complex, non-linear
- >>system (refractory periods, neurotransmitter release and uptake,...).
- >>If the brain works that way, then the simple multiply-add
- >>accumulations of backprop are over-simplified, aren't they? Then
- >>maybe both backprop and ALNs are useless for the purposes of brain
- >>modelling.
-
- >Yes!
-
- >But they are (more to the point, were) useful in showcasing that
- >one can algorithmically string together parallel computational
- >units to produce a specific behaviour. Certainly brain does not use
- >backprop or ALNs specifically, but the mathematical insight we gain
- >from them helps us to understand what's going on.
-
- Thanks for your comments. I would still like to know what the
- neurobiologists (or people who study neuron information processing,
- whoever they are) think of the idea of using a boolean logic system to
- capture the functionality of how a neuron's axon firings depend on the
- axon firings of neurons which synapse upon it. If you look at the
- little "mosquito" demo in atree 2.6, you see how ALNs are used to
- process 80-bit input vectors to learn to predict malaria (a toy
- problem, not a real one). The only difference in using atree for
- studying axon firings is that you would have to use more inputs
- (thousands) and you would have to represent signals at several input
- time steps. It would be interesting if someone had data on axonal
- firings in a collection of neurons that one could apply atree to.
-
- Any pointers to people working on this who might be interested?
-
- Bill
-
-
- --
- ***************************************************
- Prof. William W. Armstrong, Computing Science Dept.
- University of Alberta; Edmonton, Alberta, Canada T6G 2H1
- arms@cs.ualberta.ca Tel(403)492 2374 FAX 492 1071
-