<steve at tropheus.demon.co.uk> wrote:
> During my 20 years working on artificial neural networks I have become
> more and more convince that they have very little similarity to the
> real neural networks in the brain. I'm not aware of any biological
> equivalent to the weights and bias in an artificial neuron. If such
> weights exist then I would suggest that they are set and adjusted by
> the local release of neuro-chemicals rather than by some action
> similar to back propagation in artificial neural networks.
I don't know, if I have this 100% right, but I'll just give it a try.
Neural Network does not neccesarily mean artificial. Artificial networks
would be what you find in applications like databases, face-recogniton,
etc. Neuronal Networks on the other hand are computerbased simulations
of biological neurons and networks consisting of them.
In a neuron the weight of a synapse can be described through three
1)the number of neurotransmitter release-sites
2)the probability of neurotransmitter release from the pre-synapse
following an actionpotential
3)the density of receptors on the post-synapse
The synaptic weight is a construct which is directly proportional to
these three effects.
Christian Wilms (melvin at gmx.de)