Dynamics of Neural Circuits
John Hertz
Nordita, Copenhagen, Denmark
Abstract:
Neural circuitry is highly diverse, but almost all of it shares certain
universal features, notably high connectivity and competition between
excitation and inhibition. Another feature whose imortance has been
widely recognized and studied recently is synaptic adaptation: Effective
connection strengths between neurons change on the several-hundred-ms
timescale, and this additional layer of dynamics has consequences for the
network dynamics.
In this talk I will describe two general universality classes of the
network dynamics that occur as a result of excitatory-inhibitory conpetition
and show the way information may be encoded in the different cases. In the
first, neuronal excitation-inhibition loops give rise to oscillatory dynamics,
which can then be modulated by the slower synaptic dynamics. The antenna lobe,
which encodes olfactory stimuli in insects, offers an interesting example of
such a system. In the second class, exemplified by the neocortex of mammals,
separately strong excitatory and inhibitory interactions in densely-connected
networks lead to a self-consistent "balanced state" in which individual
neurons fire irregularly and information is carried by the average firing of
large populations. Synaptic adaptation can contribute to this information
transmission, making the firing rates subject more to changes in inputs than
to the input rates themselves, while generally enhancing the stability of
the balanced state of the network.