In article <1992Apr8.151118.4780 at watmath.waterloo.edu>, mwtilden at watmath.waterloo.edu (Mark W. Tilden) writes:
>> During a recent neural-networks lecture, it was pointed out that
> comp-sci types only model one type of human neuron, where animal (human)
> bodys contain no less than six different types.
This is bogus. I'd love to see answer ...
And what kind of neuron do com-sci types think their modeling, anyway?
Maybe their modeling a neural network, but even that is up for questioning.
My own experience is that the comp-sci "neural networks" are mass-parallel
type architectures, combined with feeback and weighting schemes. This
could be "inspired" by a biological neural network, but that's about it.
Of course, everyone has their own definition of a neural network. Try
putting a neuroscientist and an comp-sci AI person in the same room ...
>> Just when I thought we were making progress in neural-nets, Ma Nature
> throws us a 6-sided multiplier. Typical.
I believe its many orders larger than just 6-sided ....
Rob Butera - ECE grad student Rice University, Houston, TX rbutera at rice.edu
"This attractive lady, whom I had only recently been introduced to, dropped
into my lap. I chose not to dump her off." - Gary Hart