Fig 1. Cell activity characterization.
a) Firing rates νi of 6 selected neurons belonging to two anti-correlated assemblies, the color identifies the assembly and the colors correspond to the one used in b) for the different clusters; b) raster plot activity, the firing times are colored according to the assembly the neurons belong to; c) cross-correlation matrix C(νi, νj) of the firing rates. The neurons in panel b) and c) are clustered according to the correlation of their firing rates by employing the k-means algorithm; the clusters are ordered in terms of their average correlation (inside each cluster) from the highest to the lowest one (for more details see Methods). The firing rates are calculated over overlapping time windows of duration 1 s, the origins of successive windows are shifted by 50 ms. The system is evolved during 107 spikes, after discarding an initial transient of 105 spike events. Other parameters used in the simulation: g = 8, K = 20, N = 400, kmean = Nact/15, ΔV = 5 mV and τα = 20 ms. The number of active neurons is 370, corresponding to n* ≃ 93%.