Click on the image to see a PDF version (for zooming in)

Fig. 12.5. Effect of noise on desynchronization. A network of 180 neurons with both excitatory and inhibitory lateral connections was simulated for 500 iterations. The network was divided into two groups, as in the experiment of Figure 12.4. The first group (E1) consisted of neurons [1..22], [45..66], [89..110], and [133..154], and the second group (E2) of neurons [23..44], [67..88], [111..132], and [155..180]. The excitatory lateral connections were limited to the neurons in the same group within a radius of 90; the inhibitory connections were global. To illustrate, the plots at right show the lateral connections of neuron 45 in E1 and 132 in E2. (a) The membrane potential of each neuron was uniformly randomly initialized, and no noise was added afterward. The symmetry is broken and the two groups are separated as expected. (b) The membrane potentials initially were the same, but perturbed throughout the simulation by adding 0.1% of uniformly random noise. The neurons within the same group are synchronized at the same time as the two groups are desynchronized. (c) Without any noise (initial or continual), the symmetry was not broken and the entire network remained synchronized. A small amount of noise is therefore essential for proper desynchronization to occur.