From Working Memory to Epilepsy: Dynamics of Facilitation and Inhibition in a Cortical Network
Sergio Verduzco
Department of Psychology and Neuroscience, University of Colorado at Boulder
Date and time:
Thursday, October 6, 2011 - 4:00pm
Abstract:
Certain tasks like writing a phone number or following driving directions require a type a type of short-term memory known as “working memory.” The substrate of working memory consists of persistent activity in populations of neurons which are selective to the stimulus being remembered. Modelers use recurrently connected neural networks to replicate the persistent activity observed in real neurons, and the states of elevated firing rates constitute attractors in phase space. One property of neural networks with recurrent excitatory connections is the presence of unstable states, which raises the possibility of working memory networks being implicated in the generation of seizures. We studied the dynamics of a simplified working memory model to investigate this possibility. Using symmetry reductions and decomposition into fast and slow subsystems we characterized the pathological dynamics of working memory networks with reduced inhibition, which include generalized oscillations and frequency selectivity.