Biomedical Engineering Reference
In-Depth Information
cerned with adding inhibitory cells in our network models. Because the effects
of inhibition can depend on many factors (local versus diffuse connectivity, pha-
sic versus tonic, etc.), this will certainly require the use of cell-based ensemble
models (as by Timofeev et al. (39) and Tsodyks et al. (41)), instead of the mean-
field type of model presented here.
It will be particularly important to understand how network organization is
changed as gabaergic synapses switch from excitatory to inhibitory during de-
velopment, a transition that may be driven by network activity itself (13). This
transition occurs in parallel to developmental changes in excitatory connectivity
and cellular properties, changes which may also, in part, be due to spontaneous
activity. But how does activity modify these network properties? Our under-
standing of the spontaneous activity in developing systems will facilitate the
study of the role of this activity for network maturation. It will become neces-
sary to identify the long-term mechanisms of activity-dependent plasticity oper-
ating in developing networks. These "learning rules" will then be added to our
models of spontaneous activity, allowing us to study how activity in a network
leads to changes in that network, changes which in turn will affect activity (see
(19) for an example of how synaptic plasticity leads to changes in the patterns of
activity). This effect of activity on itself, by way of modifying network proper-
ties, is one striking feature of neural network complexity.
6.
ACKNOWLEDGMENTS
We thank Jay Demas for providing his unpublished data on mouse retinal
activity. We also thank Cristina Marchetti for her comments on the manuscript.
7.
NOTES
1. When a neuron discharges an action potential, it releases neurotransmit-
ters from its synaptic terminals onto postsynaptic neurons. Some neurotransmit-
ters will increase the membrane potential of the postsynaptic neurons, which
makes the postsynaptic neurons more likely to fire action potentials. This is
what we mean by "excitatory connection." Other transmitters will decrease post-
synaptic membrane potential or increase membrane conductance (shunt) such
that the postsynaptic neuron is less likely to generate action potentials; this type
of connection is called "inhibitory."
2. We use a "top-down," "rate-coded" approach, as defined in the previous
chapter 5.1. by Reeke.
