By Ferreira, F.; Erlhagen, W.; Bicho, E.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Recent evidence suggests that the neural mechanisms underlying memory for serial order and interval timing of sequential events are closely linked. We present a dynamic neural field model which exploits the existence and stability of multi-bump solutions with a gradient of activation to store serial order. The activation gradient is achieved by applying a state-dependent threshold accommodation process to the firing rate function. A field dynamics of lateral inhibition type is used in combination with a dynamics for the baseline activity to recall the sequence from memory. We show that depending on the time scale of the baseline dynamics the precise temporal structure of the original sequence may be retrieved or a proactive timing of events may be achieved