Ray,
Some other thoughts.
It occurs to me that one of the things we don't simulate with ANNs is
the delay in recovering fluency in the practice of a skill. Assuming
that this is encoded via some route through a network and we imagine the
route as involving the transport of a resource (call it action
potentials, if you will), then, as the resource is pushed through each
axon, some must get left behind. This acts as a repository of resource
which effectively lowers the threshhold for that axon. If the axon
doesn't fire for a while, this resource may get transformed into a more
stable substance or absorbed into some nearby intra-cellular structure.
Subsequent resumption of the skill triggers the release of this reserve
of resource, but it takes a little while, hence, we are not immediately
fluent upon the resumption of a skill abandoned for some little while.
Another thought. Does the neuron, as fed by the resource arriving via
dendrites, provide all the resource for punching through its axons? Does
the arriving resource just act as a switch, allowing resource from the
neurocellular fluid to flow into the cell and through the axon? Is this
a basis for mood-inhibition of neurocellular activity (i.e. lack of
resource in the neurocellular fluid inhibits resource flow through axons
to whatever has arrived via dendrites or is locked up in the cell). If
one neuron has many axons, and has to supply all the resource to fire
them from a smaller number of feeding dendrites, the intensity of signal
through the axons would be reduced in comparison to the arriving signal.
--
Jeff Best
jeffb at jtbest.demon.co.uk