Figure 3.

Development of synchrony in feedforward networks. (A) Spike rasters from a simulation of a randomly connected feedforward network. Each cell receives n_e = 1400 excitatory and n_i = 600 inhibitory inputs. In addition, two cells in a layer share, on average, a proportion p = 0.05 of their inputs. Each cell in layer 1 receives an independent Poisson excitatory input, so that outputs from the first layer are uncorrelated. (B) A feedforward network with no overlap. Each cell receives the same number of inputs as in (A), but there are no shared inputs (p = 0). Correlated inputs are introduced to the first layer, ${\rho }_1^{\text{in}}=0.05,$ to match the level of correlation introduced by overlap in the input to layer 2 in (A). (C) A feedforward network with no overlap receiving independent input. All model parameters are the same as in (B). However, the input to the first layer is uncorrelated $({\rho }_1^{\text{in}}=0)$, and synchrony does not develop. The spike count correlation over a window of width 50 ms averaged over all pairs is ρ = 0.02, 0.18, and 0.59 for layers 2, 3, and 4 in (A); and ρ = 0.03, 0.21, and 0.63 for layers 1, 2, and 3 in (B). Cells in all other layers are not correlated.