A neural network is presented that stores spatio-temporal patterns (synfire-chains) in associative networks of spiking neurons and replays them at a controllable speed. An implicit equation is derived and solved numerically which relates theaverage speed to the network parameters. The replay speed can be controlled by unspecific background signals and also depends on the number of co-activated synfire-chains. Balanced inhibition can prevent the latter dependency. Simulation results confirmthe theory, but reveal instabilities for low and high control signals. These boundaries are traced back to four different destabilizing mechanisms.
展开▼
