Study of noise influence on self-organization phenomena in chaotic neural networks

In our previous studies several interesting dynamic phenomena have been discovered in chaotic neural networks with third-order continuous-time neurons. The considered network had only local interactions between the nodes. Depending on the connection structure and initial conditions imposed (which can be considered as external stimulation) for each neuron the array exhibits various kinds of spatially ordered chaotic waves: target waves, checkerboard patterns, auto-wave-like behavior and spatial memory effect. The patterns with visible spatial regularity (as opposed to spatial chaos) emerging from cooperation of a large number of chaotic dynamic elements are referred to as self-organization. In this paper we analyze how inevitable noise existing in the network affects the observed phenomena. Depending on the noise level introduced we observe the effect of wave propagation failure and loss of the memory effect. For high levels of initial noise the wave phenomena are completely destroyed by noise-induced spontaneous oscillations.