Figure 7: Mechanisms of self-organized network connectivity.
From: Spontaneous emergence of fast attractor dynamics in a model of developing primary visual cortex
(left) Lateral excitatory connections make the network intrinsically unstable in the absence of inhibition. The full network, with lateral connections and mutual inhibition (red curve), produces low spontaneous firing rates. However, when inhibition is removed, but excitatory lateral connections are preserved, the firing rate diverges to a high constant value (green curve). This effect disappears when all lateral connections (both inhibitory and excitatory) are removed, restoring low firing rates (blue curve). Right: average firing rates for a 100 ms stimulus presentation, for each of four cell clusters, using the preferred stimulus for each cluster, both with the full network (red curves) and after removing all lateral connections (blue curves). Dotted vertical lines indicate stimulus offset at t=100 ms. Recurrent connectivity results in a small overall amplification, which does not noticeably slow down network dynamics. Notice that all curves quickly decay to zero after stimulus offset.
