Tempered neural networks
William Sulis · 2003
Tempered neural networks, which are cellular automaton neural networks in which the updating occurs according to the arrival times of action potentials within the network, are introduced. Simulations of integer timed networks displayed low activity levels for individual neurons, bursts consisting of the cooperative activity of groups of neurons, and encoding of information in complex spatiotemporal patterns. At low input levels a new form of memory called dynamic memory was demonstrated. A new route to chaos at high input levels is suggested.>