CORTEX conference by Henry Kennedy

Surprisingly little is known about the statistics of cortical networks due to an absence of investigation of their weighted and spatial properties. Using brain-wide retrograde tracing experiments in macaque, we are generating a consistent database of between area connections with projection densities, and distances. The network is neither a sparse small-world graph nor scale-free. Local connectivity accounts for 80% of labeled neurons, meaning that cortex is heavily involved in local function. Importantly link weights, are highly characteristic across animals, follow a heavy-tailed lognormal distribution over 6 orders of magnitude, and decay exponentially with distance. We predict that the statistical properties of the cortex will give insight into the nature of the processing mode of the cortex. We are making weighted network analysis, this reveals a trade off between local and global efficiencies. An important finding is that a distance rule predicts the binary features, the global and local communication efficiencies as well as the clustered topography of the graph. These findings underline the importance ofweight-based hierarchical layering in cortical architecture and processing.