Stability analysis of a multi-area network model of macaque visual cortex

We present our current work on a model comprising the 32 areas of the macaque cortex associatedwith visual processing, where the individual areas are based on a layered spiking network model of earlysensory cortex. Combining a simple neuron model with complex connectivity enables us to study the influ-ence of the structural connectivity itself on cortical dynamics. Finding the parameter regime that combinessufficient excitability with stable dynamics at the same time is challenging in this high dimensional andcomplex model. In this talk we present a reduction of the high-dimensional spiking model to a simplersystem, replacing the dynamical variables of each single neuron by mean firing rates of populations. Theapproach is based on mean-field techniques [1] and enables us to theoretically predict the mean activity inthe network model as well as its stability properties. We show how the mean-field theory helps us findinga good working point with asynchronous activity and realistically low firing rates. In particular we perform astability analysis near the instable fixed points of the system and deduce mathematical rules for refining thestructural connectivity in order to enlarge the basin of attraction of the fixed points with realistic activity. Wecompare our predictions to the spiking simulation and show that the modified connectivity ensures stablenetwork dynamics in the spiking model while fulfilling the biological constraints