132
spiking. What is needed is a method to identify the underrepresented branches and
then determine those that should receive inputs and how strong these inputs should
be. While this technique has yet to be tried in any rigorous setting, preliminary
results have shown that it can yield more accurate results than using just one input.
5.3 Network Simulations of Reduced Cells
A natural extension of the work in this thesis is to employ reduced cells in net-
work simulations. Prototypical reductions can be computed for cells with different
biophysical properties. For example, if a network of the hippocampus is to be mod-
eled with regions CA3 and CAl represented, each with 1000 cells in each region, then
at least two prototypes could be created: one being a reduction of a stereotypical
CA3 cell and the other of a stereotypical CAl cell. Note that these cells can have
different kinetics and morphologies, as well as different reduced system sizes. How-
ever, within each brain region there are multiple (but a finite number of) subtypes
of cells, and some are excitatory while others are inhibitory (Shepherd and Koch,
1998) (Traub and Miles, 1991). Hence the network population of 1000 cells may be
composed of, say, 20 different prototype cells, each represented according to their
proportion of occurrence in the brain. These reduced cells are then building blocks
that can be linked together to form networks by implementing a synaptic transmis-
sion mechanism. This opens the door to investigation of other neuronal functions,
such as synaptic plasticity, using realistic morphologies but without the price of slow