Banbury 2006 — Abstracts
Controlling neural network dynamics
Kanaka Rajan & Larry F. Abbott
We explore mechanisms by which a set of constant inputs to a neural network can generate predictable and repeatable patterns of activity with complex dynamics. Model networks of firing-rate or spiking neurons can exhibit dynamic activity ranging from fixed point behavior to oscillations or chaos. These patterns of activity can play useful cognitive and functional roles (Amit, 1989). Taking advantage of this richness, however, requires control of network dynamics so that a desired form of activity can be called up when it is needed. We explore mechanisms by which such control can be exerted.