Re: Making neural nets more like real neurons

From: eugen@leitl.org
Date: Mon Apr 01 2002 - 02:29:43 MST


On Mon, 1 Apr 2002, Rüdiger Koch wrote:

> Nope. But I would not worry about this part of uploading technology. Once
> scanning technology is developed enough we can scan a mouse and instantiate
> it into different neuron models until we find it eats virtual cheese ;-)

A more robust approach would seem to start with a series of
well-characterized organisms, e.g. D. discoideum -> C. elegans -> Aplysia
-> D. melanogaster->M. musculus-> H. sapiens, validating behaviour in
machina along each step.

Apart from scaling issues one would have to develop new methods at each
step. C. elegans is still small enough for a conventional manual tracing
from EM micrographs, but the next steps will require machine vision, and
then automated imaging. Mice (and certainly men) are too large to be
imaged en bloc (the surface, that is), requiring recursive subdision into
individually imaged blocks.

Another issue is development of new simulation paradigms, as larger
organisms are impractical to simulate even on large scale molecular
circuitry machines.



This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:13:11 MST